@wlearn/ensemble 0.1.0 → 0.2.0

package/CHANGELOG.md

@@ -0,0 +1,17 @@
+# Changelog
+
+## 0.2.0
+
+- Depend on the CommonJS `@wlearn/core` release and remove the unused direct `@wlearn/types` dependency
+- Add package homepage and GitHub issue metadata
+
+## 0.1.0
+
+- Initial release
+- VotingEnsemble: soft/hard voting with configurable weights
+- StackingEnsemble: K-fold stacking with customizable meta-learner
+- BaggedEstimator: bootstrap aggregating with configurable sample size
+- Caruana greedy ensemble selection
+- OOF prediction utilities
+- Weight optimization with simplex projection
+- 60 tests

package/LICENSE

@@ -0,0 +1,177 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to the Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by the Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding any notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS

package/NOTICE

@@ -0,0 +1,5 @@
+@wlearn/ensemble
+
+Copyright (c) 2026 Anton Zemlyansky and contributors
+
+This package is part of the wlearn project and is licensed under the Apache License, Version 2.0.

package/README.md

@@ -0,0 +1,92 @@
+# @wlearn/ensemble
+
+Ensemble methods for wlearn: voting, stacking, bagging, and Caruana greedy selection.
+
+Part of [wlearn](https://wlearn.org) ([GitHub](https://github.com/wlearn-org), [all packages](https://github.com/wlearn-org/wlearn#repository-structure)).
+
+## Install
+
+```bash
+npm install @wlearn/ensemble
+```
+
+## Voting
+
+Combine multiple models by averaging predictions (soft) or majority vote (hard).
+
+```js
+const { VotingEnsemble } = require('@wlearn/ensemble')
+const { LinearModel } = require('@wlearn/liblinear')
+const { XGBModel } = require('@wlearn/xgboost')
+
+const ens = await VotingEnsemble.create({
+  estimators: [
+    ['linear', LinearModel, { task: 'classification' }],
+    ['xgb', XGBModel, { task: 'classification' }]
+  ],
+  voting: 'soft',
+  task: 'classification'
+})
+
+ens.fit(X, y)
+const preds = ens.predict(X_test)
+const acc = ens.score(X_test, y_test)
+
+ens.dispose()
+```
+
+## Stacking
+
+Train base models, collect out-of-fold predictions, then train a meta-learner on those predictions.
+
+```js
+const { StackingEnsemble } = require('@wlearn/ensemble')
+const { LinearModel } = require('@wlearn/liblinear')
+const { XGBModel } = require('@wlearn/xgboost')
+
+const stack = await StackingEnsemble.create({
+  estimators: [
+    ['xgb', XGBModel, { task: 'classification' }],
+    ['linear', LinearModel, { task: 'classification' }]
+  ],
+  finalEstimator: ['meta', LinearModel, { task: 'classification' }],
+  cv: 5,
+  task: 'classification'
+})
+
+stack.fit(X, y)
+const preds = stack.predict(X_test)
+
+stack.dispose()
+```
+
+## Bagging
+
+Bootstrap aggregating: train multiple copies of the same model on bootstrap samples.
+
+```js
+const { BaggedEstimator } = require('@wlearn/ensemble')
+const { LinearModel } = require('@wlearn/liblinear')
+
+const bag = await BaggedEstimator.create({
+  estimator: ['linear', LinearModel, { task: 'classification' }],
+  nEstimators: 10,
+  task: 'classification'
+})
+
+bag.fit(X, y)
+const preds = bag.predict(X_test)
+
+bag.dispose()
+```
+
+## Utilities
+
+- `caruanaSelect(oofPredictions, yTrue, opts?)` -- Caruana greedy ensemble selection
+- `getOofPredictions(estimatorSpecs, X, y, opts?)` -- compute out-of-fold predictions
+- `optimizeWeights(oofPredictions, yTrue, opts?)` -- optimize ensemble weights
+- `projectSimplex(weights)` -- project weights onto probability simplex
+
+## License
+
+Apache-2.0

package/dist/ensemble.js

@@ -0,0 +1 @@
+var ensemble=(()=>{var x=(s,t)=>()=>(t||s((t={exports:{}}).exports,t),t.exports);var k=x((Yr,kt)=>{var O=class extends Error{constructor(t,n){super(t),this.name="WlearnError",this.code=n||"ERR_WLEARN"}},rt=class extends O{constructor(t){super(t,"ERR_BUNDLE"),this.name="BundleError"}},ot=class extends O{constructor(t){super(t,"ERR_REGISTRY"),this.name="RegistryError"}},it=class extends O{constructor(t){super(t,"ERR_VALIDATION"),this.name="ValidationError"}},at=class extends O{constructor(t="Model is not fitted. Call fit() first."){super(t,"ERR_NOT_FITTED"),this.name="NotFittedError"}},ct=class extends O{constructor(t="Model has been disposed."){super(t,"ERR_DISPOSED"),this.name="DisposedError"}};kt.exports={WlearnError:O,BundleError:rt,RegistryError:ot,ValidationError:it,NotFittedError:at,DisposedError:ct}});var U=x((Tr,Mt)=>{var{ValidationError:M}=k();function as(s,t="auto"){if(s&&typeof s=="object"&&!Array.isArray(s)&&s.data){let{data:n,rows:e,cols:o}=s;if(!(n instanceof Float64Array)){if(t==="error")throw new M("Expected Float64Array in typed matrix");return{data:new Float64Array(n),rows:e,cols:o}}return{data:n,rows:e,cols:o}}if(Array.isArray(s)&&Array.isArray(s[0])){if(t==="error")throw new M('Input coercion disabled (coerce: "error"). Pass { data: Float64Array, rows, cols } instead of number[][].');let n=s.length,e=s[0].length,o=new Float64Array(n*e);for(let r=0;r<n;r++)for(let i=0;i<e;i++)o[r*e+i]=s[r][i];if(t==="warn"){let r=o.byteLength;console.warn(`@wlearn/core: Converted number[][] to Float64Array (copied ${(r/1024).toFixed(1)} KB, shape ${n}x${e}). For performance, pass { data, rows, cols }.`)}return{data:o,rows:n,cols:e}}throw new M("X must be number[][] or { data: Float64Array, rows, cols }")}function cs(s){return s instanceof Int32Array||s instanceof Float32Array||s instanceof Float64Array?s:new Float64Array(s)}function ls(s,t,n){if(!t||!n||t<1||n<1)throw new M(`Invalid dimensions: rows=${t}, cols=${n}`);if(!(s instanceof Float32Array)&&!(s instanceof Float64Array))throw new M("data must be Float32Array or Float64Array");if(s.length!==t*n)throw new M(`data.length (${s.length}) !== rows * cols (${t*n})`);return{data:s,rows:t,cols:n}}function fs(s){if(!s||typeof s!="object")throw new M("Matrix must be an object");let{data:t,rows:n,cols:e}=s;if(typeof n!="number"||typeof e!="number"||n<1||e<1)throw new M(`Invalid dimensions: rows=${n}, cols=${e}`);if(!(t instanceof Float32Array)&&!(t instanceof Float64Array))throw new M("data must be Float32Array or Float64Array");if(t.length!==n*e)throw new M(`data.length (${t.length}) !== rows * cols (${n*e})`);return s}Mt.exports={normalizeX:as,normalizeY:cs,makeDense:ls,validateMatrix:fs}});var lt=x((Cr,qt)=>{var hs=new Uint32Array([1116352408,1899447441,3049323471,3921009573,961987163,1508970993,2453635748,2870763221,3624381080,310598401,607225278,1426881987,1925078388,2162078206,2614888103,3248222580,3835390401,4022224774,264347078,604807628,770255983,1249150122,1555081692,1996064986,2554220882,2821834349,2952996808,3210313671,3336571891,3584528711,113926993,338241895,666307205,773529912,1294757372,1396182291,1695183700,1986661051,2177026350,2456956037,2730485921,2820302411,3259730800,3345764771,3516065817,3600352804,4094571909,275423344,430227734,506948616,659060556,883997877,958139571,1322822218,1537002063,1747873779,1955562222,2024104815,2227730452,2361852424,2428436474,2756734187,3204031479,3329325298]),N=new Uint32Array(64);function q(s,t){return s>>>t|s<<32-t}function us(s,t,n){for(let l=0;l<16;l++)N[l]=(t[n]<<24|t[n+1]<<16|t[n+2]<<8|t[n+3])>>>0,n+=4;for(let l=16;l<64;l++){let d=q(N[l-15],7)^q(N[l-15],18)^N[l-15]>>>3,u=q(N[l-2],17)^q(N[l-2],19)^N[l-2]>>>10;N[l]=N[l-16]+d+N[l-7]+u>>>0}let e=s[0],o=s[1],r=s[2],i=s[3],a=s[4],f=s[5],h=s[6],c=s[7];for(let l=0;l<64;l++){let d=q(a,6)^q(a,11)^q(a,25),u=a&f^~a&h,p=c+d+u+hs[l]+N[l]>>>0,m=q(e,2)^q(e,13)^q(e,22),w=e&o^e&r^o&r,b=m+w>>>0;c=h,h=f,f=a,a=i+p>>>0,i=r,r=o,o=e,e=p+b>>>0}s[0]=s[0]+e>>>0,s[1]=s[1]+o>>>0,s[2]=s[2]+r>>>0,s[3]=s[3]+i>>>0,s[4]=s[4]+a>>>0,s[5]=s[5]+f>>>0,s[6]=s[6]+h>>>0,s[7]=s[7]+c>>>0}function ds(s){let t=s instanceof Uint8Array?s:new Uint8Array(s),n=t.length,e=n*8,o=n+9+63&-64,r=new Uint8Array(o);r.set(t),r[n]=128;let i=new DataView(r.buffer);i.setUint32(o-4,e>>>0,!1),e>4294967295&&i.setUint32(o-8,e/4294967296>>>0,!1);let a=new Uint32Array([1779033703,3144134277,1013904242,2773480762,1359893119,2600822924,528734635,1541459225]);for(let h=0;h<o;h+=64)us(a,r,h);let f="";for(let h=0;h<8;h++)f+=a[h].toString(16).padStart(8,"0");return f}qt.exports={sha256Sync:ds}});var $t=x((Wr,_t)=>{var ps=new Uint8Array([87,76,82,78]),ms=1,ws=16,gs={FLOAT32:"float32",FLOAT64:"float64",INT32:"int32"};_t.exports={BUNDLE_MAGIC:ps,BUNDLE_VERSION:ms,HEADER_SIZE:ws,DTYPE:gs}});var C=x((Jr,Dt)=>{var{BUNDLE_MAGIC:Ot,BUNDLE_VERSION:G,HEADER_SIZE:I}=$t(),{BundleError:$}=k(),{sha256Sync:Lt}=lt();function T(s){return s==null||typeof s!="object"?JSON.stringify(s):Array.isArray(s)?"["+s.map(n=>T(n)).join(",")+"]":"{"+Object.keys(s).sort().map(n=>JSON.stringify(n)+":"+T(s[n])).join(",")+"}"}var ft=new TextEncoder,ht=new TextDecoder;function bs(s,t){let n=[...t].sort((u,p)=>u.id<p.id?-1:u.id>p.id?1:0),e=0,o=[],r=[];for(let u of n){let p=u.data instanceof Uint8Array?u.data:new Uint8Array(u.data),m=Lt(p),w={id:u.id,offset:e,length:p.length,sha256:m};u.mediaType&&(w.mediaType=u.mediaType),o.push(w),r.push(p),e+=p.length}let i={...s,bundleVersion:G},a=ft.encode(T(i)),f=ft.encode(T(o)),h=I+a.length+f.length+e,c=new Uint8Array(h),l=new DataView(c.buffer);c.set(Ot,0),l.setUint32(4,G,!0),l.setUint32(8,a.length,!0),l.setUint32(12,f.length,!0),c.set(a,I),c.set(f,I+a.length);let d=I+a.length+f.length;for(let u of r)c.set(u,d),d+=u.length;return c}function Bt(s){let t=s instanceof Uint8Array?s:new Uint8Array(s);if(t.length<I)throw new $(`Bundle too small: ${t.length} bytes (minimum ${I})`);for(let l=0;l<4;l++)if(t[l]!==Ot[l])throw new $("Invalid bundle magic (expected WLRN)");let n=new DataView(t.buffer,t.byteOffset,t.byteLength),e=n.getUint32(4,!0);if(e!==G)throw new $(`Unsupported bundle version: ${e} (expected ${G})`);let o=n.getUint32(8,!0),r=n.getUint32(12,!0);if(I+o+r>t.length)throw new $(`Bundle truncated: header declares ${I+o+r} bytes but got ${t.length}`);let i,a;try{i=JSON.parse(ht.decode(t.subarray(I,I+o)))}catch(l){throw new $(`Invalid manifest JSON: ${l.message}`)}try{a=JSON.parse(ht.decode(t.subarray(I+o,I+o+r)))}catch(l){throw new $(`Invalid TOC JSON: ${l.message}`)}let f=I+o+r,h=t.length-f;for(let l=0;l<a.length;l++){let d=a[l];if(d.offset<0||d.length<0||d.offset+d.length>h)throw new $(`TOC entry "${d.id}" out of bounds: offset=${d.offset}, length=${d.length}, blobRegion=${h}`);for(let u=l+1;u<a.length;u++){let p=a[u],m=d.offset,w=d.offset+d.length,b=p.offset,g=p.offset+p.length;if(m<g&&b<w&&d.length>0&&p.length>0)throw new $(`TOC entries "${d.id}" and "${p.id}" overlap`)}}let c=t.subarray(f);return{manifest:i,toc:a,blobs:c}}function ys(s){return ft.encode(T(s))}function xs(s){let t=s instanceof Uint8Array?s:new Uint8Array(s);return JSON.parse(ht.decode(t))}function As(s){let{manifest:t,toc:n,blobs:e}=Bt(s);for(let o of n){let r=e.subarray(o.offset,o.offset+o.length),i=Lt(r);if(i!==o.sha256)throw new $(`SHA-256 mismatch for "${o.id}": expected ${o.sha256}, got ${i}`)}return{manifest:t,toc:n,blobs:e}}Dt.exports={encodeBundle:bs,decodeBundle:Bt,encodeJSON:ys,decodeJSON:xs,validateBundle:As}});var K=x((Gr,Rt)=>{var{RegistryError:D}=k(),{decodeBundle:zt}=C(),V=new Map;function Fs(s,t){if(typeof s!="string"||!s.includes("@"))throw new D(`Invalid typeId "${s}": must contain "@" (e.g. "wlearn.liblinear.classifier@1")`);if(typeof t!="function")throw new D("loaderFn must be a function");V.set(s,t)}async function Ss(s){let{manifest:t,toc:n,blobs:e}=zt(s),{typeId:o}=t;if(!o)throw new D("Bundle manifest missing typeId");let r=V.get(o);if(!r){let i=[...V.keys()],a=i.length>0?`Registered loaders: ${i.join(", ")}`:"No loaders registered";throw new D(`No loader registered for typeId "${o}". ${a}. Install the corresponding @wlearn/* package and import it to register the loader.`)}return await r(t,n,e)}function Es(s){let{manifest:t,toc:n,blobs:e}=zt(s),{typeId:o}=t;if(!o)throw new D("Bundle manifest missing typeId");let r=V.get(o);if(!r){let a=[...V.keys()],f=a.length>0?`Registered loaders: ${a.join(", ")}`:"No loaders registered";throw new D(`No loader registered for typeId "${o}". ${f}. Install the corresponding @wlearn/* package and import it to register the loader.`)}let i=r(t,n,e);if(i&&typeof i.then=="function")throw new D(`Loader for "${o}" returned a Promise. Use async load() instead of loadSync().`);return i}function Is(){return new Map(V)}Rt.exports={register:Fs,load:Ss,loadSync:Es,getRegistry:Is}});var dt=x((Kr,jt)=>{var ut=class{#t;#e;constructor(t,n){this.#t=t,this.#e=n}get name(){return this.#t}get estimator(){return this.#e}get isFitted(){return this.#e.isFitted}get isTransformer(){return typeof this.#e.transform=="function"}};jt.exports={Step:ut}});var Tt=x((Hr,Yt)=>{var{Step:Ps}=dt(),{DisposedError:vs,NotFittedError:Ns,ValidationError:H}=k(),{encodeBundle:ks,decodeBundle:Ms}=C(),{register:qs,load:Ut}=K(),Vt="wlearn.pipeline@1",Xt=!1,Z=class s{#t;#e=!1;#r=!1;constructor(t){if(this.#t=t.map(([n,e])=>new Ps(n,e)),this.#t.length===0)throw new H("Pipeline requires at least one step")}#n(){if(this.#r)throw new vs("Pipeline has been disposed.")}#o(){if(this.#n(),!this.#e)throw new Ns("Pipeline is not fitted. Call fit() first.")}#s(t){let n=t;for(let e=0;e<this.#t.length-1;e++)n=this.#t[e].estimator.transform(n);return n}fit(t,n){this.#n();let e=t;for(let r=0;r<this.#t.length-1;r++){let i=this.#t[r].estimator;typeof i.fitTransform=="function"?e=i.fitTransform(e,n):(i.fit(e,n),e=i.transform(e))}return this.#t[this.#t.length-1].estimator.fit(e,n),this.#e=!0,this}predict(t){this.#o();let n=this.#s(t);return this.#t[this.#t.length-1].estimator.predict(n)}predictProba(t){this.#o();let n=this.#t[this.#t.length-1].estimator;if(typeof n.predictProba!="function")throw new H("Last step does not support predictProba");let e=this.#s(t);return n.predictProba(e)}score(t,n){this.#o();let e=this.#s(t);return this.#t[this.#t.length-1].estimator.score(e,n)}save(){this.#o();let t={typeId:Vt,steps:this.#t.map(e=>({name:e.name,params:e.estimator.getParams()}))},n=this.#t.map(e=>({id:e.name,data:e.estimator.save(),mediaType:"application/x-wlearn-bundle"}));return ks(t,n)}static async load(t){let{manifest:n,toc:e,blobs:o}=Ms(t),r=[];for(let a of n.steps){let f=e.find(l=>l.id===a.name);if(!f)throw new H(`No artifact found for pipeline step "${a.name}"`);let h=o.subarray(f.offset,f.offset+f.length),c=await Ut(h);r.push([a.name,c])}let i=new s(r);return i.#e=!0,i}dispose(){if(!this.#r){this.#r=!0;for(let t of this.#t)t.estimator.dispose()}}getParams(){let t={};for(let n of this.#t)t[n.name]=n.estimator.getParams();return t}setParams(t){this.#n();for(let n of this.#t)t[n.name]&&n.estimator.setParams(t[n.name]);return this}get capabilities(){return this.#t[this.#t.length-1].estimator.capabilities}get isFitted(){return this.#e}static registerLoader(){Xt||(Xt=!0,qs(Vt,(t,n,e)=>s._loadFromParts(t,n,e)))}static async _loadFromParts(t,n,e){let o=[];for(let i of t.steps){let a=n.find(c=>c.id===i.name);if(!a)throw new H(`No artifact found for pipeline step "${i.name}"`);let f=e.subarray(a.offset,a.offset+a.length),h=await Ut(f);o.push([i.name,h])}let r=new s(o);return r.#e=!0,r}};Z.registerLoader();Yt.exports={Pipeline:Z}});var Gt=x((Qr,Jt)=>{var{normalizeX:Ct,normalizeY:Zr}=U(),{ValidationError:Wt}=k(),pt=class s{#t;#e=!1;#r=null;#n=null;#o=null;#s=null;#i=0;constructor(t={}){this.#t={impute:t.impute??"auto",encode:t.encode??"auto",scale:t.scale??!1,maxCategories:t.maxCategories??20,...t}}fit(t,n){let e=Ct(t),{data:o,rows:r,cols:i}=e;this.#r=new Array(i),this.#n=new Float64Array(i);let a=new Array(i);for(let h=0;h<i;h++){let c=[],l=!1,d=0,u=0;for(let b=0;b<r;b++){let g=o[b*i+h];Number.isNaN(g)?l=!0:(c.push(g),d+=g,u++)}let p=new Set(c),m=c.every(b=>b===Math.floor(b)),w=this.#t.encode!==!1&&m&&p.size<=this.#t.maxCategories&&p.size>=2;if(this.#r[h]=w?"categorical":"numeric",this.#t.impute!==!1&&l)if(w){let b=new Map;for(let F of c)b.set(F,(b.get(F)||0)+1);let g=c[0],y=0;for(let[F,E]of b)E>y&&(g=F,y=E);this.#n[h]=g}else this.#n[h]=u>0?d/u:0;a[h]={values:c,unique:p,sum:d,count:u}}this.#o=new Array(i).fill(null);let f=0;if(this.#t.encode!==!1)for(let h=0;h<i;h++)if(this.#r[h]==="categorical"){let c=[...a[h].unique].sort((l,d)=>l-d);if(this.#t.encode==="label"||this.#t.encode===!1){let l=new Map;c.forEach((d,u)=>l.set(d,u)),this.#o[h]={type:"label",mapping:l,startIdx:f},f++}else{let l=new Map;c.forEach((d,u)=>l.set(d,f+u)),this.#o[h]={type:"onehot",mapping:l,startIdx:f,size:c.length},f+=c.length}}else f++;else f=i;if(this.#i=f,this.#s=new Array(this.#i).fill(null),this.#t.scale){let h=0;for(let c=0;c<i;c++){if(this.#r[c]==="categorical"&&this.#t.encode!==!1){let d=this.#o[c];h+=d.type==="onehot"?d.size:1;continue}let l=a[c].values;if(this.#t.scale==="standard"){let d=a[c].sum/(a[c].count||1),u=0;for(let m of l)u+=(m-d)**2;u/=l.length||1;let p=Math.sqrt(u);this.#s[h]={mean:d,std:p||1}}else if(this.#t.scale==="minmax"){let d=1/0,u=-1/0;for(let m of l)m<d&&(d=m),m>u&&(u=m);let p=u-d||1;this.#s[h]={min:d,range:p}}h++}}return this.#e=!0,this}transform(t){if(!this.#e)throw new Wt("Preprocessor not fitted");let n=Ct(t),{data:e,rows:o,cols:r}=n,i=this.#i,a=new Float64Array(o*i);for(let f=0;f<o;f++){let h=0;for(let c=0;c<r;c++){let l=e[f*r+c];if(Number.isNaN(l)&&this.#t.impute!==!1&&(l=this.#n[c]),this.#r[c]==="categorical"&&this.#o[c]){let d=this.#o[c];if(d.type==="onehot"){let u=d.mapping.get(l);u!==void 0&&(a[f*i+u]=1),h=d.startIdx+d.size}else{let u=d.mapping.get(l);a[f*i+h]=u!==void 0?u:-1,h++}}else a[f*i+h]=l,h++}if(this.#t.scale)for(let c=0;c<i;c++){let l=this.#s[c];l&&(this.#t.scale==="standard"?a[f*i+c]=(a[f*i+c]-l.mean)/l.std:this.#t.scale==="minmax"&&(a[f*i+c]=(a[f*i+c]-l.min)/l.range))}}return{data:a,rows:o,cols:i}}fitTransform(t,n){return this.fit(t,n),this.transform(t)}getState(){if(!this.#e)throw new Wt("Preprocessor not fitted");return{config:this.#t,colTypes:this.#r,imputeValues:[...this.#n],encodings:this.#o.map(t=>t?{type:t.type,mapping:[...t.mapping.entries()],startIdx:t.startIdx,size:t.size}:null),scaleParams:this.#s,outputCols:this.#i}}static fromState(t){let n=new s(t.config);return n.#r=t.colTypes,n.#n=new Float64Array(t.imputeValues),n.#o=t.encodings.map(e=>e?{type:e.type,mapping:new Map(e.mapping),startIdx:e.startIdx,size:e.size}:null),n.#s=t.scaleParams,n.#i=t.outputCols,n.#e=!0,n}get isFitted(){return this.#e}get outputCols(){return this.#i}get capabilities(){return{transformer:!0}}dispose(){}};Jt.exports={Preprocessor:pt}});var oe=x((to,re)=>{var{NotFittedError:Kt,DisposedError:Ht,ValidationError:X}=k(),{normalizeX:Q}=U(),{encodeBundle:Zt,encodeJSON:Qt,decodeJSON:te}=C(),{register:ee}=K(),se="wlearn.preprocess.standard_scaler@1",ne="wlearn.preprocess.minmax_scaler@1",tt=class s{#t=null;#e=null;#r=!1;#n=!1;#o={};constructor(t={}){this.#o={...t}}fit(t){this.#s();let{rows:n,cols:e,data:o}=Q(t);if(n===0)throw new X("Cannot fit on empty data");let r=new Float64Array(e),i=new Float64Array(e);for(let f=0;f<n;f++){let h=f+1;for(let c=0;c<e;c++){let l=o[f*e+c],d=l-r[c];r[c]+=d/h;let u=l-r[c];i[c]+=d*u}}let a=new Float64Array(e);for(let f=0;f<e;f++)a[f]=n>1?Math.sqrt(i[f]/(n-1)):0;return this.#t=r,this.#e=a,this.#r=!0,this}transform(t){this.#i();let{rows:n,cols:e,data:o}=Q(t);if(e!==this.#t.length)throw new X(`Expected ${this.#t.length} columns, got ${e}`);let r=new Float64Array(n*e);for(let i=0;i<n;i++)for(let a=0;a<e;a++){let f=i*e+a,h=this.#e[a];r[f]=h>0?(o[f]-this.#t[a])/h:0}return{rows:n,cols:e,data:r}}fitTransform(t){return this.fit(t),this.transform(t)}save(){this.#i();let t={means:Array.from(this.#t),stds:Array.from(this.#e)};return Zt({typeId:se,params:this.getParams()},[{id:"params",data:Qt(t),mediaType:"application/json"}])}static _fromBundle(t,n,e){let o=n.find(a=>a.id==="params");if(!o)throw new X('Bundle missing "params" artifact');let r=te(e.subarray(o.offset,o.offset+o.length)),i=new s(t.params||{});return i.#t=new Float64Array(r.means),i.#e=new Float64Array(r.stds),i.#r=!0,i}dispose(){this.#n||(this.#n=!0,this.#t=null,this.#e=null,this.#r=!1)}getParams(){return{...this.#o}}setParams(t){return Object.assign(this.#o,t),this}get isFitted(){return this.#r&&!this.#n}#s(){if(this.#n)throw new Ht("StandardScaler has been disposed.")}#i(){if(this.#s(),!this.#r)throw new Kt("StandardScaler is not fitted.")}},et=class s{#t=null;#e=null;#r=!1;#n=!1;#o={};constructor(t={}){this.#o={...t}}fit(t){this.#s();let{rows:n,cols:e,data:o}=Q(t);if(n===0)throw new X("Cannot fit on empty data");let r=new Float64Array(e).fill(1/0),i=new Float64Array(e).fill(-1/0);for(let a=0;a<n;a++)for(let f=0;f<e;f++){let h=o[a*e+f];h<r[f]&&(r[f]=h),h>i[f]&&(i[f]=h)}return this.#t=r,this.#e=i,this.#r=!0,this}transform(t){this.#i();let{rows:n,cols:e,data:o}=Q(t);if(e!==this.#t.length)throw new X(`Expected ${this.#t.length} columns, got ${e}`);let r=new Float64Array(n*e);for(let i=0;i<n;i++)for(let a=0;a<e;a++){let f=i*e+a,h=this.#e[a]-this.#t[a];r[f]=h>0?(o[f]-this.#t[a])/h:0}return{rows:n,cols:e,data:r}}fitTransform(t){return this.fit(t),this.transform(t)}save(){this.#i();let t={mins:Array.from(this.#t),maxs:Array.from(this.#e)};return Zt({typeId:ne,params:this.getParams()},[{id:"params",data:Qt(t),mediaType:"application/json"}])}static _fromBundle(t,n,e){let o=n.find(a=>a.id==="params");if(!o)throw new X('Bundle missing "params" artifact');let r=te(e.subarray(o.offset,o.offset+o.length)),i=new s(t.params||{});return i.#t=new Float64Array(r.mins),i.#e=new Float64Array(r.maxs),i.#r=!0,i}dispose(){this.#n||(this.#n=!0,this.#t=null,this.#e=null,this.#r=!1)}getParams(){return{...this.#o}}setParams(t){return Object.assign(this.#o,t),this}get isFitted(){return this.#r&&!this.#n}#s(){if(this.#n)throw new Ht("MinMaxScaler has been disposed.")}#i(){if(this.#s(),!this.#r)throw new Kt("MinMaxScaler is not fitted.")}};ee(se,tt._fromBundle);ee(ne,et._fromBundle);re.exports={StandardScaler:tt,MinMaxScaler:et}});var mt=x((eo,ie)=>{function _s(s=42){let t=s|0;return()=>(t=t*1664525+1013904223&2147483647,t/2147483647)}function $s(s,t){for(let n=s.length-1;n>0;n--){let e=t()*(n+1)|0,o=s[n];s[n]=s[e],s[e]=o}return s}ie.exports={makeLCG:_s,shuffle:$s}});var bt=x((so,ae)=>{var{ValidationError:L}=k();function B(s,t,n){if(!s||!t||s.length===0||t.length===0)throw new L(`${n}: inputs must be non-empty`);if(s.length!==t.length)throw new L(`${n}: length mismatch (${s.length} vs ${t.length})`)}function Y(s,t){let n=new Set;for(let r=0;r<s.length;r++)n.add(s[r]);if(t)for(let r=0;r<t.length;r++)n.add(t[r]);let e=new Int32Array([...n].sort((r,i)=>r-i)),o=new Map;for(let r=0;r<e.length;r++)o.set(e[r],r);return{labels:e,labelMap:o,nClasses:e.length}}function st(s,t,n,e){let o=new Int32Array(e*e);for(let r=0;r<s.length;r++){let i=n.get(s[r]),a=n.get(t[r]);o[i*e+a]++}return o}function wt(s,t){let n=new Float64Array(t),e=new Float64Array(t),o=new Float64Array(t);for(let r=0;r<t;r++){n[r]=s[r*t+r];for(let i=0;i<t;i++)i!==r&&(e[r]+=s[i*t+r],o[r]+=s[r*t+i])}return{tp:n,fp:e,fn:o}}function Os(s,t){B(s,t,"accuracy");let n=0;for(let e=0;e<s.length;e++)s[e]===t[e]&&n++;return n/s.length}function Ls(s,t){B(s,t,"r2Score");let n=s.length,e=0;for(let i=0;i<n;i++)e+=s[i];e/=n;let o=0,r=0;for(let i=0;i<n;i++){let a=s[i]-e;o+=a*a;let f=s[i]-t[i];r+=f*f}return o===0?0:1-r/o}function Bs(s,t){B(s,t,"meanSquaredError");let n=0;for(let e=0;e<s.length;e++){let o=s[e]-t[e];n+=o*o}return n/s.length}function Ds(s,t){B(s,t,"meanAbsoluteError");let n=0;for(let e=0;e<s.length;e++)n+=Math.abs(s[e]-t[e]);return n/s.length}function zs(s,t){B(s,t,"confusionMatrix");let{labels:n,labelMap:e,nClasses:o}=Y(s,t);return{matrix:st(s,t,e,o),labels:n}}function gt(s,t,n){if(s==="binary"){if(t>2)throw new L('average="binary" requires exactly 2 classes');return"binary"}return s||"binary"}function Rs(s,t,{average:n="binary"}={}){B(s,t,"precisionScore");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fp:f}=wt(i,r),h=gt(n,r,e);if(h==="binary"){let l=r-1,d=a[l]+f[l];return d===0?0:a[l]/d}if(h==="micro"){let l=0,d=0;for(let u=0;u<r;u++)l+=a[u],d+=f[u];return l+d===0?0:l/(l+d)}let c=0;for(let l=0;l<r;l++){let d=a[l]+f[l];c+=d===0?0:a[l]/d}return c/r}function js(s,t,{average:n="binary"}={}){B(s,t,"recallScore");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fn:f}=wt(i,r),h=gt(n,r,e);if(h==="binary"){let l=r-1,d=a[l]+f[l];return d===0?0:a[l]/d}if(h==="micro"){let l=0,d=0;for(let u=0;u<r;u++)l+=a[u],d+=f[u];return l+d===0?0:l/(l+d)}let c=0;for(let l=0;l<r;l++){let d=a[l]+f[l];c+=d===0?0:a[l]/d}return c/r}function Us(s,t,{average:n="binary"}={}){B(s,t,"f1Score");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fp:f,fn:h}=wt(i,r),c=gt(n,r,e);function l(u,p,m){let w=u+p===0?0:u/(u+p),b=u+m===0?0:u/(u+m);return w+b===0?0:2*w*b/(w+b)}if(c==="binary"){let u=r-1;return l(a[u],f[u],h[u])}if(c==="micro"){let u=0,p=0,m=0;for(let w=0;w<r;w++)u+=a[w],p+=f[w],m+=h[w];return l(u,p,m)}let d=0;for(let u=0;u<r;u++)d+=l(a[u],f[u],h[u]);return d/r}function Vs(s,t,{nClasses:n,eps:e=1e-15}={}){if(!s||s.length===0)throw new L("logLoss: yTrue must be non-empty");let o=s.length;if(!n){let a=new Set;for(let f=0;f<o;f++)a.add(s[f]);n=a.size}if(t.length!==o*n)throw new L(`logLoss: yProba length (${t.length}) must be n * nClasses (${o*n})`);let{labelMap:r}=Y(s),i=0;for(let a=0;a<o;a++){let f=r.get(s[a]),h=t[a*n+f];h=Math.max(e,Math.min(1-e,h)),i-=Math.log(h)}return i/o}function Xs(s,t){if(!s||s.length===0)throw new L("rocAuc: yTrue must be non-empty");if(s.length!==t.length)throw new L("rocAuc: length mismatch");let{labels:n}=Y(s);if(n.length!==2)throw new L("rocAuc: requires exactly 2 classes");let e=n[n.length-1],o=s.length,r=Array.from({length:o},(u,p)=>p);r.sort((u,p)=>t[p]-t[u]);let i=0,a=0;for(let u=0;u<o;u++)s[u]===e?i++:a++;if(i===0||a===0)return 0;let f=0,h=0,c=0,l=0,d=0;for(let u=0;u<o;u++){let p=r[u];s[p]===e?f++:h++;let m=f/i,w=h/a;d+=(w-l)*(m+c)/2,c=m,l=w}return d}ae.exports={accuracy:Os,r2Score:Ls,meanSquaredError:Bs,meanAbsoluteError:Ds,confusionMatrix:zs,precisionScore:Rs,recallScore:js,f1Score:Us,logLoss:Vs,rocAuc:Xs}});var pe=x((no,de)=>{var{ValidationError:z}=k(),{makeLCG:yt,shuffle:xt}=mt(),{normalizeX:Ys,normalizeY:Ts}=U(),{accuracy:Cs,r2Score:Ws,meanSquaredError:Js,meanAbsoluteError:Gs}=bt(),ce={accuracy:(s,t)=>Cs(s,t),r2:(s,t)=>Ws(s,t),neg_mse:(s,t)=>-Js(s,t),neg_mae:(s,t)=>-Gs(s,t)};function he(s){if(typeof s=="function")return s;let t=ce[s];if(!t)throw new z(`Unknown scoring: "${s}". Available: ${Object.keys(ce).join(", ")}`);return t}function Ks(s,t=5,{shuffle:n=!0,seed:e=42}={}){if(s<t)throw new z(`kFold: n (${s}) must be >= k (${t})`);if(t<2)throw new z("kFold: k must be >= 2");let o=Int32Array.from({length:s},(h,c)=>c);if(n){let h=yt(e);xt(o,h)}let r=Math.floor(s/t),i=s%t,a=[],f=0;for(let h=0;h<t;h++){let c=r+(h<i?1:0),l=o.slice(f,f+c),d=[];f>0&&d.push(o.slice(0,f)),f+c<s&&d.push(o.slice(f+c));let u=Qs(d);a.push({train:u,test:l}),f+=c}return a}function ue(s,t=5,{shuffle:n=!0,seed:e=42}={}){let o=s.length;if(o<t)throw new z(`stratifiedKFold: n (${o}) must be >= k (${t})`);if(t<2)throw new z("stratifiedKFold: k must be >= 2");let r=new Map;for(let h=0;h<o;h++){let c=s[h];r.has(c)||r.set(c,[]),r.get(c).push(h)}if(n){let h=yt(e);for(let c of r.values())xt(c,h)}let i=Array.from({length:t},()=>[]);for(let h of r.values())for(let c=0;c<h.length;c++)i[c%t].push(h[c]);let a=Int32Array.from({length:o},(h,c)=>c),f=[];for(let h=0;h<t;h++){let c=new Set(i[h]),l=new Int32Array(i[h]),d=a.filter(u=>!c.has(u));f.push({train:d,test:l})}return f}function Hs(s,{testSize:t=.2,shuffle:n=!0,seed:e=42}={}){if(s<2)throw new z("trainTestSplit: n must be >= 2");let o=Math.max(1,Math.round(s*t)),r=s-o;if(r<1)throw new z("trainTestSplit: testSize too large");let i=Int32Array.from({length:s},(a,f)=>f);if(n){let a=yt(e);xt(i,a)}return{train:i.slice(0,r),test:i.slice(r)}}async function Zs(s,t,n,{cv:e=5,scoring:o="accuracy",seed:r=42,params:i={}}={}){let a=Ys(t),f=Ts(n),h=he(o),c;Array.isArray(e)?c=e:c=ue(f,e,{shuffle:!0,seed:r});let l=new Float64Array(c.length);for(let d=0;d<c.length;d++){let{train:u,test:p}=c[d],m=le(a,u),w=fe(f,u),b=le(a,p),g=fe(f,p),y=await s.create(i);try{y.fit(m,w);let F=await y.predict(b);l[d]=h(g,F)}finally{y.dispose()}}return l}function Qs(s){let t=0;for(let o of s)t+=o.length;let n=new Int32Array(t),e=0;for(let o of s)n.set(o,e),e+=o.length;return n}function le(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function fe(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}de.exports={kFold:Ks,stratifiedKFold:ue,trainTestSplit:Hs,crossValScore:Zs,getScorer:he}});var ge=x((ro,we)=>{function me(s){return s!=null&&typeof s.then=="function"}function tn(s,t){return me(s)?s.then(t):t(s)}we.exports={isPromiseLike:me,lift:tn}});var Ae=x((oo,xe)=>{var{normalizeY:en}=U(),{round:sn}=Math;function be(s){let t=en(s);if(t instanceof Int32Array)return"classification";let n=new Set;for(let e=0;e<t.length;e++){if(t[e]!==sn(t[e]))return"regression";n.add(t[e])}return n.size<=20?"classification":"regression"}var ye=new WeakMap;function P(s){let t=ye.get(s);if(!t)throw new Error("Model: invalid instance");return t}function W(s,t){let n=P(s);if(n.disposed)throw new Error(`${t} has been disposed.`);if(!n.inner)throw new Error(`${t}: not fitted`);return n.inner}function nn(s,t,n={}){let e=n.name||"Model",o=n.load||null,r=s===t;class i{constructor(c,l){ye.set(this,{inner:null,task:c,params:l,disposed:!1})}static async create(c={}){let l={...c},d=l.task||null;delete l.task,o&&await o();let u=new i(d,l),p=P(u);if(d){let m=d==="classification"?s:t;p.inner=await m.create({...l,task:d})}else r&&(p.inner=await s.create(l));return u}static async load(c){try{let l=await s.load(c),d=new i("classification",{}),u=P(d);return u.inner=l,typeof l.getParams=="function"&&(u.params=l.getParams(),u.params.task&&(u.task=u.params.task)),d}catch{let d=await t.load(c),u=new i("regression",{}),p=P(u);return p.inner=d,typeof d.getParams=="function"&&(p.params=d.getParams(),p.params.task&&(p.task=p.params.task)),u}}fit(c,l,d){let u=P(this);if(u.task||(u.task=be(l)),!u.inner){let p=u.task==="classification"?s:t;u.inner=new p({...u.params,task:u.task})}return u.inner.fit(c,l,d),this}predict(c,l){return W(this,e).predict(c,l)}predictProba(c,l){let d=W(this,e);if(typeof d.predictProba!="function")throw new Error(`${e}: predictProba not available`);return d.predictProba(c,l)}score(c,l,d){return W(this,e).score(c,l,d)}save(){return W(this,e).save()}dispose(){let c=P(this);c.inner&&typeof c.inner.dispose=="function"&&c.inner.dispose(),c.inner=null,c.disposed=!0}getParams(){let c=P(this),l=c.inner&&typeof c.inner.getParams=="function"?c.inner.getParams():{...c.params};return c.task&&(l.task=c.task),l}setParams(c){let l=P(this),d={...c};if(d.task){let u=d.task;delete d.task,u!==l.task&&l.inner&&this.dispose(),l.task=u}return l.params=d,l.inner&&typeof l.inner.setParams=="function"&&l.inner.setParams(d),this}get task(){return P(this).task}get isFitted(){let c=P(this);return c.inner?c.inner.isFitted!==void 0?c.inner.isFitted:!0:!1}get classes(){let c=P(this);return c.inner?typeof c.inner.classes=="function"?c.inner.classes():c.inner.classes!==void 0?c.inner.classes:new Int32Array(0):new Int32Array(0)}get capabilities(){let c=P(this);return c.inner&&c.inner.capabilities!==void 0?c.inner.capabilities:{}}}let a=new Set(["constructor","fit","predict","predictProba","score","save","dispose","getParams","setParams"]),f=new Set(["classes","task","isFitted","capabilities"]);for(let h of[s,t])if(!(!h||!h.prototype))for(let c of Object.getOwnPropertyNames(h.prototype)){if(c.startsWith("_")||c.startsWith("#")||a.has(c)||f.has(c)||c in i.prototype)continue;let l=Object.getOwnPropertyDescriptor(h.prototype,c);l&&(typeof l.value=="function"?i.prototype[c]=function(...d){return W(this,e)[c](...d)}:l.get&&Object.defineProperty(i.prototype,c,{get(){let d=P(this);if(d.inner)return d.inner[c]},configurable:!0}))}return(s.defaultSearchSpace||t.defaultSearchSpace)&&(i.defaultSearchSpace=()=>s.defaultSearchSpace?.()||t.defaultSearchSpace?.()||{}),(s.budgetSpec||t.budgetSpec)&&(i.budgetSpec=()=>s.budgetSpec?.()||t.budgetSpec?.()||void 0),Object.defineProperty(i,"name",{value:e,configurable:!0}),i}xe.exports={createModelClass:nn,detectTask:be}});var R=x((io,Fe)=>{var{WlearnError:rn,BundleError:on,RegistryError:an,ValidationError:cn,NotFittedError:ln,DisposedError:fn}=k(),{normalizeX:hn,normalizeY:un,makeDense:dn,validateMatrix:pn}=U(),{sha256Sync:mn}=lt(),{encodeBundle:wn,decodeBundle:gn,validateBundle:bn,encodeJSON:yn,decodeJSON:xn}=C(),{register:An,load:Fn,loadSync:Sn,getRegistry:En}=K(),{Pipeline:In}=Tt(),{Step:Pn}=dt(),{Preprocessor:vn}=Gt(),{StandardScaler:Nn,MinMaxScaler:kn}=oe(),{makeLCG:Mn,shuffle:qn}=mt(),{accuracy:_n,r2Score:$n,meanSquaredError:On,meanAbsoluteError:Ln,confusionMatrix:Bn,precisionScore:Dn,recallScore:zn,f1Score:Rn,logLoss:jn,rocAuc:Un}=bt(),{kFold:Vn,stratifiedKFold:Xn,trainTestSplit:Yn,crossValScore:Tn,getScorer:Cn}=pe(),{isPromiseLike:Wn,lift:Jn}=ge(),{createModelClass:Gn,detectTask:Kn}=Ae();Fe.exports={WlearnError:rn,BundleError:on,RegistryError:an,ValidationError:cn,NotFittedError:ln,DisposedError:fn,normalizeX:hn,normalizeY:un,makeDense:dn,validateMatrix:pn,sha256Sync:mn,encodeBundle:wn,decodeBundle:gn,validateBundle:bn,encodeJSON:yn,decodeJSON:xn,register:An,load:Fn,loadSync:Sn,getRegistry:En,Pipeline:In,Step:Pn,Preprocessor:vn,StandardScaler:Nn,MinMaxScaler:kn,makeLCG:Mn,shuffle:qn,accuracy:_n,r2Score:$n,meanSquaredError:On,meanAbsoluteError:Ln,confusionMatrix:Bn,precisionScore:Dn,recallScore:zn,f1Score:Rn,logLoss:jn,rocAuc:Un,kFold:Vn,stratifiedKFold:Xn,trainTestSplit:Yn,crossValScore:Tn,getScorer:Cn,isPromiseLike:Wn,lift:Jn,createModelClass:Gn,detectTask:Kn}});var qe=x((ao,Me)=>{var{encodeBundle:Hn,decodeBundle:Zn,register:Se,load:Ee,normalizeX:At,normalizeY:Ie,accuracy:Qn,r2Score:tr,ValidationError:nt,NotFittedError:er,DisposedError:sr,lift:Pe}=R(),ve="wlearn.ensemble.voting.classifier@1",Ne="wlearn.ensemble.voting.regressor@1",ke=!1,Ft=class s{#t;#e;#r;#n;#o;#s;#i=!1;#f=!1;constructor(t){this.#t=t.estimators||[],this.#r=t.weights||null,this.#n=t.voting||"soft",this.#o=t.task||"classification",this.#e=null,this.#s=null,s._register()}static async create(t={}){return new s(t)}#l(){if(this.#f)throw new sr("VotingEnsemble has been disposed.")}#a(){if(this.#l(),!this.#i)throw new er("VotingEnsemble is not fitted. Call fit() first.")}async fit(t,n){this.#l();let e=At(t),o=Ie(n);if(this.#o==="classification"){let r=new Set;for(let i=0;i<o.length;i++)r.add(o[i]);this.#s=new Int32Array([...r].sort((i,a)=>i-a))}this.#r||(this.#r=new Float64Array(this.#t.length).fill(1/this.#t.length)),this.#e=[];for(let[r,i,a]of this.#t){let f=await i.create(a||{});f.fit(e,o),this.#e.push(f)}return this.#i=!0,this}predict(t){this.#a();let n=At(t),e=n.rows;if(this.#o==="regression")return this.#c(n,e);if(this.#n==="soft"){let o=this.predictProba(n);return Pe(o,r=>{let i=this.#s.length,a=new Float64Array(e);for(let f=0;f<e;f++){let h=0,c=-1/0;for(let l=0;l<i;l++)r[f*i+l]>c&&(c=r[f*i+l],h=l);a[f]=this.#s[h]}return a})}return this.#h(n,e)}predictProba(t){if(this.#a(),this.#o!=="classification")throw new nt("predictProba is only available for classification");if(this.#n==="hard")throw new nt('predictProba requires voting="soft"');let n=At(t),e=n.rows,o=this.#s.length,r=[],i=!1;for(let f=0;f<this.#e.length;f++){let h=this.#e[f].predictProba(n);h!=null&&typeof h.then=="function"&&(i=!0),r.push(h)}let a=f=>{let h=new Float64Array(e*o);for(let c=0;c<f.length;c++){let l=f[c],d=this.#r[c];for(let u=0;u<e*o;u++)h[u]+=d*l[u]}return h};return i?Promise.all(r).then(a):a(r)}score(t,n){this.#a();let e=this.predict(t),o=Ie(n),r=this.#o==="classification"?Qn:tr;return Pe(e,i=>r(o,i))}save(){this.#a();let n={typeId:this.#o==="classification"?ve:Ne,params:{task:this.#o,voting:this.#n,weights:[...this.#r],estimatorNames:this.#t.map(o=>o[0]),classes:this.#s?[...this.#s]:null}},e=this.#e.map((o,r)=>({id:this.#t[r][0],data:o.save(),mediaType:"application/x-wlearn-bundle"}));return Hn(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=Zn(t),r=n.params,i=new s({task:r.task,voting:r.voting,weights:new Float64Array(r.weights)});i.#s=r.classes?new Int32Array(r.classes):null,i.#t=r.estimatorNames.map(a=>[a,null,null]),i.#e=[];for(let a of r.estimatorNames){let f=e.find(l=>l.id===a);if(!f)throw new nt(`No artifact for estimator "${a}"`);let h=o.subarray(f.offset,f.offset+f.length),c=await Ee(h);i.#e.push(c)}return i.#i=!0,i}dispose(){if(!this.#f&&(this.#f=!0,this.#e))for(let t of this.#e)t.dispose()}getParams(){return{task:this.#o,voting:this.#n,weights:this.#r?[...this.#r]:null,estimatorNames:this.#t.map(t=>t[0])}}setParams(t){return this.#l(),t.voting!==void 0&&(this.#n=t.voting),t.weights!==void 0&&(this.#r=new Float64Array(t.weights)),this}get capabilities(){return{classifier:this.#o==="classification",regressor:this.#o==="regression",predictProba:this.#o==="classification"&&this.#n==="soft",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#i}get classes(){return this.#s}#c(t,n){let e=[],o=!1;for(let i=0;i<this.#e.length;i++){let a=this.#e[i].predict(t);a!=null&&typeof a.then=="function"&&(o=!0),e.push(a)}let r=i=>{let a=new Float64Array(n);for(let f=0;f<i.length;f++){let h=this.#r[f];for(let c=0;c<n;c++)a[c]+=h*i[f][c]}return a};return o?Promise.all(e).then(r):r(e)}#h(t,n){let e=[],o=!1;for(let i=0;i<this.#e.length;i++){let a=this.#e[i].predict(t);a!=null&&typeof a.then=="function"&&(o=!0),e.push(a)}let r=i=>{let a=this.#s.length,f=new Float64Array(n);for(let h=0;h<n;h++){let c=new Float64Array(a);for(let u=0;u<i.length;u++){let p=i[u][h],m=this.#s.indexOf(p);m>=0&&(c[m]+=this.#r[u])}let l=0,d=-1/0;for(let u=0;u<a;u++)c[u]>d&&(d=c[u],l=u);f[h]=this.#s[l]}return f};return o?Promise.all(e).then(r):r(e)}static _register(){if(ke)return;ke=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);Se(ve,t),Se(Ne,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,voting:o.voting,weights:new Float64Array(o.weights)});r.#s=o.classes?new Int32Array(o.classes):null,r.#t=o.estimatorNames.map(i=>[i,null,null]),r.#e=[];for(let i of o.estimatorNames){let a=n.find(c=>c.id===i);if(!a)throw new nt(`No artifact for estimator "${i}"`);let f=e.subarray(a.offset,a.offset+a.length),h=await Ee(f);r.#e.push(h)}return r.#i=!0,r}};Me.exports={VotingEnsemble:Ft}});var Ue=x((co,je)=>{var{encodeBundle:nr,decodeBundle:rr,register:_e,load:$e,normalizeX:Oe,normalizeY:Le,accuracy:or,r2Score:ir,stratifiedKFold:ar,kFold:cr,ValidationError:J,NotFittedError:lr,DisposedError:fr,lift:St}=R(),Be="wlearn.ensemble.stacking.classifier@1",De="wlearn.ensemble.stacking.regressor@1",ze=!1,Et=class s{#t;#e;#r;#n;#o;#s;#i;#f;#l;#a;#c;#h=!1;#d=!1;constructor(t){this.#t=t.estimators||[],this.#e=t.finalEstimator||null,this.#o=t.cv||5,this.#s=t.task||"classification",this.#i=t.passthrough||!1,this.#f=t.seed??42,this.#r=null,this.#n=null,this.#l=null,this.#a=0,this.#c=0,s._register()}static async create(t={}){return new s(t)}#p(){if(this.#d)throw new fr("StackingEnsemble has been disposed.")}#u(){if(this.#p(),!this.#h)throw new lr("StackingEnsemble is not fitted. Call fit() first.")}async fit(t,n){if(this.#p(),!this.#e)throw new J("StackingEnsemble requires a finalEstimator");let e=Oe(t),o=Le(n),r=e.rows;if(this.#s==="classification"){let p=new Set;for(let m=0;m<o.length;m++)p.add(o[m]);this.#l=new Int32Array([...p].sort((m,w)=>m-w)),this.#a=this.#l.length}let i=this.#s==="classification"?ar(o,this.#o,{shuffle:!0,seed:this.#f}):cr(r,this.#o,{shuffle:!0,seed:this.#f}),a=this.#t.length,f=this.#s==="classification"?this.#a:1,h=a*f,c=new Float64Array(r*h);for(let p=0;p<a;p++){let[,m,w]=this.#t[p];for(let{train:b,test:g}of i){let y=Re(e,b),F=hr(o,b),E=Re(e,g),v=await m.create(w||{});try{if(v.fit(y,F),this.#s==="classification"){let S=await v.predictProba(E);for(let A=0;A<g.length;A++){let j=g[A];for(let _=0;_<this.#a;_++)c[j*h+p*f+_]=S[A*this.#a+_]}}else{let S=await v.predict(E);for(let A=0;A<g.length;A++)c[g[A]*h+p]=S[A]}}finally{v.dispose()}}}let l;if(this.#i){this.#c=h+e.cols;let p=new Float64Array(r*this.#c);for(let m=0;m<r;m++)p.set(c.subarray(m*h,(m+1)*h),m*this.#c),p.set(e.data.subarray(m*e.cols,(m+1)*e.cols),m*this.#c+h);l={data:p,rows:r,cols:this.#c}}else this.#c=h,l={data:c,rows:r,cols:h};this.#r=[];for(let[,p,m]of this.#t){let w=await p.create(m||{});w.fit(e,o),this.#r.push(w)}let[,d,u]=this.#e;return this.#n=await d.create(u||{}),this.#n.fit(l,o),this.#h=!0,this}predict(t){this.#u();let n=this.#m(t);return St(n,e=>this.#n.predict(e))}predictProba(t){if(this.#u(),this.#s!=="classification")throw new J("predictProba is only available for classification");if(typeof this.#n.predictProba!="function")throw new J("Meta-model does not support predictProba");let n=this.#m(t);return St(n,e=>this.#n.predictProba(e))}score(t,n){this.#u();let e=this.predict(t),o=Le(n),r=this.#s==="classification"?or:ir;return St(e,i=>r(o,i))}save(){this.#u();let n={typeId:this.#s==="classification"?Be:De,params:{task:this.#s,cv:this.#o,passthrough:this.#i,seed:this.#f,estimatorNames:this.#t.map(o=>o[0]),metaName:this.#e[0],classes:this.#l?[...this.#l]:null,nMetaCols:this.#c}},e=this.#r.map((o,r)=>({id:this.#t[r][0],data:o.save(),mediaType:"application/x-wlearn-bundle"}));return e.push({id:this.#e[0],data:this.#n.save(),mediaType:"application/x-wlearn-bundle"}),nr(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=rr(t);return s._loadFromParts(n,e,o)}dispose(){if(!this.#d){if(this.#d=!0,this.#r)for(let t of this.#r)t.dispose();this.#n&&this.#n.dispose()}}getParams(){return{task:this.#s,cv:this.#o,passthrough:this.#i,seed:this.#f,estimatorNames:this.#t.map(t=>t[0]),metaName:this.#e?this.#e[0]:null}}setParams(t){return this.#p(),t.cv!==void 0&&(this.#o=t.cv),t.passthrough!==void 0&&(this.#i=t.passthrough),t.seed!==void 0&&(this.#f=t.seed),this}get capabilities(){return{classifier:this.#s==="classification",regressor:this.#s==="regression",predictProba:this.#s==="classification",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#h}get classes(){return this.#l}#m(t){let n=Oe(t),e=n.rows,o=this.#r.length,r=this.#s==="classification"?this.#a:1,i=o*r,a=[],f=!1;for(let c=0;c<o;c++){let l=this.#s==="classification"?this.#r[c].predictProba(n):this.#r[c].predict(n);l!=null&&typeof l.then=="function"&&(f=!0),a.push(l)}let h=c=>{let l=new Float64Array(e*this.#c);for(let d=0;d<o;d++)if(this.#s==="classification"){let u=c[d];for(let p=0;p<e;p++)for(let m=0;m<this.#a;m++)l[p*this.#c+d*r+m]=u[p*this.#a+m]}else{let u=c[d];for(let p=0;p<e;p++)l[p*this.#c+d]=u[p]}if(this.#i)for(let d=0;d<e;d++)l.set(n.data.subarray(d*n.cols,(d+1)*n.cols),d*this.#c+i);return{data:l,rows:e,cols:this.#c}};return f?Promise.all(a).then(h):h(a)}static _register(){if(ze)return;ze=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);_e(Be,t),_e(De,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,cv:o.cv,passthrough:o.passthrough,seed:o.seed});r.#l=o.classes?new Int32Array(o.classes):null,r.#a=r.#l?r.#l.length:0,r.#c=o.nMetaCols,r.#t=o.estimatorNames.map(f=>[f,null,null]),r.#e=[o.metaName,null,null],r.#r=[];for(let f of o.estimatorNames){let h=n.find(l=>l.id===f);if(!h)throw new J(`No artifact for base estimator "${f}"`);let c=e.subarray(h.offset,h.offset+h.length);r.#r.push(await $e(c))}let i=n.find(f=>f.id===o.metaName);if(!i)throw new J(`No artifact for meta estimator "${o.metaName}"`);let a=e.subarray(i.offset,i.offset+i.length);return r.#n=await $e(a),r.#h=!0,r}};function Re(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function hr(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}je.exports={StackingEnsemble:Et}});var He=x((lo,Ke)=>{var{encodeBundle:ur,decodeBundle:dr,register:Ve,load:pr,normalizeX:It,normalizeY:Xe,accuracy:mr,r2Score:wr,stratifiedKFold:gr,kFold:br,ValidationError:Ye,NotFittedError:yr,DisposedError:xr,lift:Te}=R(),Ce="wlearn.ensemble.bagged.classifier@1",We="wlearn.ensemble.bagged.regressor@1",Je=!1,Pt=class s{#t;#e;#r;#n;#o;#s;#i;#f=0;#l=0;#a;#c;#h=!1;#d=!1;constructor(t={}){this.#t=t.estimator||null,this.#e=t.kFold||5,this.#r=t.nRepeats||1,this.#n=t.task||"classification",this.#o=t.seed??42,this.#s=null,this.#i=null,this.#a=null,this.#c=null,s._register()}static async create(t={}){return new s(t)}#p(){if(this.#d)throw new xr("BaggedEstimator has been disposed.")}#u(){if(this.#p(),!this.#h)throw new yr("BaggedEstimator is not fitted. Call fit() first.")}async fit(t,n){this.#p();let e=It(t),o=Xe(n),r=e.rows;if(this.#l=r,this.#n==="classification"){let f=new Set;for(let h=0;h<o.length;h++)f.add(o[h]);this.#i=new Int32Array([...f].sort((h,c)=>h-c)),this.#f=this.#i.length}this.#n==="classification"?this.#a=new Float64Array(r*this.#f):this.#a=new Float64Array(r),this.#c=new Uint8Array(r);let[,i,a]=this.#t;this.#s=[];for(let f=0;f<this.#r;f++){let h=this.#o+f,c=this.#n==="classification"?gr(o,this.#e,{shuffle:!0,seed:h}):br(r,this.#e,{shuffle:!0,seed:h});for(let{train:l,test:d}of c){let u=Ge(e,l),p=Ar(o,l),m=Ge(e,d),w=await i.create(a||{});if(w.fit(u,p),this.#n==="classification"){let b=await w.predictProba(m),g=this.#f;for(let y=0;y<d.length;y++){let F=d[y];for(let E=0;E<g;E++)this.#a[F*g+E]+=b[y*g+E]}}else{let b=await w.predict(m);for(let g=0;g<d.length;g++)this.#a[d[g]]+=b[g]}for(let b=0;b<d.length;b++)this.#c[d[b]]+=1;this.#s.push(w)}}return this.#h=!0,this}predict(t){this.#u();let n=It(t),e=n.rows;if(this.#n==="regression")return this.#m(n,e);let o=this.predictProba(n);return Te(o,r=>{let i=this.#f,a=new Float64Array(e);for(let f=0;f<e;f++){let h=0,c=-1/0;for(let l=0;l<i;l++)r[f*i+l]>c&&(c=r[f*i+l],h=l);a[f]=this.#i[h]}return a})}predictProba(t){if(this.#u(),this.#n!=="classification")throw new Ye("predictProba is only available for classification");let n=It(t),e=n.rows,o=this.#f,r=this.#s.length,i=[],a=!1;for(let h of this.#s){let c=h.predictProba(n);c!=null&&typeof c.then=="function"&&(a=!0),i.push(c)}let f=h=>{let c=new Float64Array(e*o);for(let l of h)for(let d=0;d<e*o;d++)c[d]+=l[d];for(let l=0;l<e*o;l++)c[l]/=r;return c};return a?Promise.all(i).then(f):f(i)}score(t,n){this.#u();let e=this.predict(t),o=Xe(n),r=this.#n==="classification"?mr:wr;return Te(e,i=>r(o,i))}#m(t,n){let e=this.#s.length,o=[],r=!1;for(let a of this.#s){let f=a.predict(t);f!=null&&typeof f.then=="function"&&(r=!0),o.push(f)}let i=a=>{let f=new Float64Array(n);for(let h of a)for(let c=0;c<n;c++)f[c]+=h[c];for(let h=0;h<n;h++)f[h]/=e;return f};return r?Promise.all(o).then(i):i(o)}get oofPredictions(){this.#u();let t=new Uint8Array(this.#c);for(let e=0;e<t.length;e++)t[e]===0&&(t[e]=1);if(this.#n==="classification"){let e=this.#f,o=new Float64Array(this.#a);for(let r=0;r<this.#l;r++){let i=t[r];for(let a=0;a<e;a++)o[r*e+a]/=i}return o}let n=new Float64Array(this.#a);for(let e=0;e<this.#l;e++)n[e]/=t[e];return n}save(){this.#u();let n={typeId:this.#n==="classification"?Ce:We,params:{task:this.#n,kFold:this.#e,nRepeats:this.#r,seed:this.#o,estimatorName:this.#t[0],classes:this.#i?[...this.#i]:null,nClasses:this.#f,nSamples:this.#l}},e=this.#s.map((i,a)=>({id:`fold_${a}`,data:i.save(),mediaType:"application/x-wlearn-bundle"})),o=this.oofPredictions,r=new Uint8Array(o.buffer,o.byteOffset,o.byteLength);return e.push({id:"oof",data:r,mediaType:"application/octet-stream"}),ur(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=dr(t);return s._loadFromParts(n,e,o)}dispose(){if(!this.#d){if(this.#d=!0,this.#s)for(let t of this.#s)t.dispose();this.#s=null,this.#a=null,this.#c=null}}getParams(){return{task:this.#n,kFold:this.#e,nRepeats:this.#r,seed:this.#o,estimatorName:this.#t?this.#t[0]:null}}setParams(t){return this.#p(),t.kFold!==void 0&&(this.#e=t.kFold),t.nRepeats!==void 0&&(this.#r=t.nRepeats),t.seed!==void 0&&(this.#o=t.seed),this}get capabilities(){return{classifier:this.#n==="classification",regressor:this.#n==="regression",predictProba:this.#n==="classification",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#h&&!this.#d}get classes(){return this.#i}static _register(){if(Je)return;Je=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);Ve(Ce,t),Ve(We,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,kFold:o.kFold||5,nRepeats:o.nRepeats||1,seed:o.seed??42});r.#i=o.classes?new Int32Array(o.classes):null,r.#f=o.nClasses||0,r.#l=o.nSamples||0,r.#t=[o.estimatorName||"base",null,null];let i=r.#e*r.#r;r.#s=[];for(let f=0;f<i;f++){let h=`fold_${f}`,c=n.find(d=>d.id===h);if(!c)throw new Ye(`No artifact for "${h}"`);let l=e.subarray(c.offset,c.offset+c.length);r.#s.push(await pr(l))}let a=n.find(f=>f.id==="oof");if(a){let f=e.subarray(a.offset,a.offset+a.length),h=new Float64Array(f.buffer.slice(f.byteOffset,f.byteOffset+f.byteLength));r.#a=h,r.#c=new Uint8Array(r.#l).fill(1)}else r.#n==="classification"?r.#a=new Float64Array(r.#l*r.#f):r.#a=new Float64Array(r.#l),r.#c=new Uint8Array(r.#l);return r.#h=!0,r}};function Ge(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function Ar(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}Ke.exports={BaggedEstimator:Pt}});var Nt=x((fo,Qe)=>{var{ValidationError:Ze}=R();function vt(s){let t=s.length;if(t===0)return new Float64Array(0);if(t===1)return new Float64Array([1]);let n=new Float64Array(s);n.sort(),n.reverse();let e=new Float64Array(t);e[0]=n[0];for(let a=1;a<t;a++)e[a]=e[a-1]+n[a];let o=-1;for(let a=0;a<t;a++)n[a]*(a+1)>e[a]-1&&(o=a);if(o<0){let a=new Float64Array(t);return a.fill(1/t),a}let r=(e[o]-1)/(o+1),i=new Float64Array(t);for(let a=0;a<t;a++)i[a]=Math.max(s[a]-r,0);return i}function Fr(s,t,n,{task:e="classification",lr:o=.05,nIter:r=100}={}){let i=s.length,a=t.length;if(i===0)throw new Ze("optimizeWeights: need at least 1 model");if(i===1)return new Float64Array([1]);let f=new Float64Array(n),h=1e-15;if(e==="classification"){let l=s[0].length/a;if(l!==Math.floor(l))throw new Ze("optimizeWeights: prediction length must be divisible by n");for(let d=0;d<r;d++){let u=new Float64Array(i);for(let m=0;m<a;m++){let w=t[m]|0,b=0;for(let g=0;g<i;g++)b+=f[g]*s[g][m*l+w];b=Math.max(b,h);for(let g=0;g<i;g++)u[g]-=s[g][m*l+w]/b}for(let m=0;m<i;m++)u[m]/=a;for(let m=0;m<i;m++)f[m]-=o*u[m];let p=vt(f);for(let m=0;m<i;m++)f[m]=p[m]}}else for(let c=0;c<r;c++){let l=new Float64Array(i);for(let u=0;u<a;u++){let p=0;for(let w=0;w<i;w++)p+=f[w]*s[w][u];let m=t[u]-p;for(let w=0;w<i;w++)l[w]-=2*m*s[w][u]}for(let u=0;u<i;u++)l[u]/=a;for(let u=0;u<i;u++)f[u]-=o*l[u];let d=vt(f);for(let u=0;u<i;u++)f[u]=d[u]}return f}Qe.exports={optimizeWeights:Fr,projectSimplex:vt}});var ss=x((ho,es)=>{var{getScorer:Sr,normalizeY:Er,ValidationError:ts}=R(),{optimizeWeights:Ir}=Nt();function Pr(s,t,{maxSize:n=20,scoring:e="accuracy",task:o="classification",nClasses:r=0,refineWeights:i=!0}={}){let a=Er(t),f=a.length,h=s.length;if(h===0)throw new ts("caruanaSelect: need at least 1 candidate");let c=Sr(e),l=s[0].length/f;if(l!==Math.floor(l))throw new ts("caruanaSelect: oofPredictions[0].length must be divisible by n");o==="classification"&&r===0&&(r=l);let d=new Float64Array(s[0].length),u=[],p=[];for(let y=0;y<n;y++){let F=-1,E=-1/0;for(let S=0;S<h;S++){let A=vr(d,s[S],y,y+1),j=Nr(A,a,c,o,r,f);j>E&&(E=j,F=S)}u.push(F),p.push(E);let v=s[F];for(let S=0;S<d.length;S++)d[S]=(y*d[S]+v[S])/(y+1)}let m=new Map;for(let y of u)m.set(y,(m.get(y)||0)+1);let w=new Int32Array([...m.keys()].sort((y,F)=>y-F)),b=new Float64Array(w.length);for(let y=0;y<w.length;y++)b[y]=m.get(w[y])/n;let g={indices:w,weights:b,scores:new Float64Array(p)};if(i&&w.length>1){let y=Array.from(w,F=>s[F]);g.weights=Ir(y,a,b,{task:o})}return g}function vr(s,t,n,e){let o=new Float64Array(s.length);for(let r=0;r<s.length;r++)o[r]=(n*s[r]+t[r])/e;return o}function Nr(s,t,n,e,o,r){if(e==="regression")return n(t,s);let i=new Float64Array(r);for(let a=0;a<r;a++){let f=0,h=-1/0;for(let c=0;c<o;c++)s[a*o+c]>h&&(h=s[a*o+c],f=c);i[a]=f}return n(t,i)}es.exports={caruanaSelect:Pr}});var os=x((po,rs)=>{var{stratifiedKFold:kr,kFold:Mr,normalizeX:qr,normalizeY:_r,ValidationError:uo}=R();async function $r(s,t,n,{cv:e=5,seed:o=42,task:r="classification"}={}){let i=qr(t),a=_r(n),f=i.rows,h=r==="classification"?kr(a,e,{shuffle:!0,seed:o}):Mr(f,e,{shuffle:!0,seed:o}),c=null,l=0;if(r==="classification"){let u=new Set;for(let p=0;p<a.length;p++)u.add(a[p]);c=new Int32Array([...u].sort((p,m)=>p-m)),l=c.length}let d=[];for(let[u,p,m]of s){let w;r==="classification"?w=new Float64Array(f*l):w=new Float64Array(f);for(let{train:b,test:g}of h){let y=ns(i,b),F=Or(a,b),E=ns(i,g),v=await p.create(m||{});try{if(v.fit(y,F),r==="classification"){let S=await v.predictProba(E);for(let A=0;A<g.length;A++){let j=g[A];for(let _=0;_<l;_++)w[j*l+_]=S[A*l+_]}}else{let S=await v.predict(E);for(let A=0;A<g.length;A++)w[g[A]]=S[A]}}finally{v.dispose()}}d.push(w)}return{oofPreds:d,classes:c}}function ns(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function Or(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}rs.exports={getOofPredictions:$r}});var Vr=x((mo,is)=>{var{VotingEnsemble:Lr}=qe(),{StackingEnsemble:Br}=Ue(),{BaggedEstimator:Dr}=He(),{caruanaSelect:zr}=ss(),{getOofPredictions:Rr}=os(),{optimizeWeights:jr,projectSimplex:Ur}=Nt();is.exports={VotingEnsemble:Lr,StackingEnsemble:Br,BaggedEstimator:Dr,caruanaSelect:zr,getOofPredictions:Rr,optimizeWeights:jr,projectSimplex:Ur}});return Vr();})();

package/dist/ensemble.mjs

@@ -0,0 +1,2 @@
+var __ensemble=(()=>{var x=(s,t)=>()=>(t||s((t={exports:{}}).exports,t),t.exports);var k=x((Yr,kt)=>{var O=class extends Error{constructor(t,n){super(t),this.name="WlearnError",this.code=n||"ERR_WLEARN"}},rt=class extends O{constructor(t){super(t,"ERR_BUNDLE"),this.name="BundleError"}},ot=class extends O{constructor(t){super(t,"ERR_REGISTRY"),this.name="RegistryError"}},it=class extends O{constructor(t){super(t,"ERR_VALIDATION"),this.name="ValidationError"}},at=class extends O{constructor(t="Model is not fitted. Call fit() first."){super(t,"ERR_NOT_FITTED"),this.name="NotFittedError"}},ct=class extends O{constructor(t="Model has been disposed."){super(t,"ERR_DISPOSED"),this.name="DisposedError"}};kt.exports={WlearnError:O,BundleError:rt,RegistryError:ot,ValidationError:it,NotFittedError:at,DisposedError:ct}});var U=x((Tr,Mt)=>{var{ValidationError:M}=k();function as(s,t="auto"){if(s&&typeof s=="object"&&!Array.isArray(s)&&s.data){let{data:n,rows:e,cols:o}=s;if(!(n instanceof Float64Array)){if(t==="error")throw new M("Expected Float64Array in typed matrix");return{data:new Float64Array(n),rows:e,cols:o}}return{data:n,rows:e,cols:o}}if(Array.isArray(s)&&Array.isArray(s[0])){if(t==="error")throw new M('Input coercion disabled (coerce: "error"). Pass { data: Float64Array, rows, cols } instead of number[][].');let n=s.length,e=s[0].length,o=new Float64Array(n*e);for(let r=0;r<n;r++)for(let i=0;i<e;i++)o[r*e+i]=s[r][i];if(t==="warn"){let r=o.byteLength;console.warn(`@wlearn/core: Converted number[][] to Float64Array (copied ${(r/1024).toFixed(1)} KB, shape ${n}x${e}). For performance, pass { data, rows, cols }.`)}return{data:o,rows:n,cols:e}}throw new M("X must be number[][] or { data: Float64Array, rows, cols }")}function cs(s){return s instanceof Int32Array||s instanceof Float32Array||s instanceof Float64Array?s:new Float64Array(s)}function ls(s,t,n){if(!t||!n||t<1||n<1)throw new M(`Invalid dimensions: rows=${t}, cols=${n}`);if(!(s instanceof Float32Array)&&!(s instanceof Float64Array))throw new M("data must be Float32Array or Float64Array");if(s.length!==t*n)throw new M(`data.length (${s.length}) !== rows * cols (${t*n})`);return{data:s,rows:t,cols:n}}function fs(s){if(!s||typeof s!="object")throw new M("Matrix must be an object");let{data:t,rows:n,cols:e}=s;if(typeof n!="number"||typeof e!="number"||n<1||e<1)throw new M(`Invalid dimensions: rows=${n}, cols=${e}`);if(!(t instanceof Float32Array)&&!(t instanceof Float64Array))throw new M("data must be Float32Array or Float64Array");if(t.length!==n*e)throw new M(`data.length (${t.length}) !== rows * cols (${n*e})`);return s}Mt.exports={normalizeX:as,normalizeY:cs,makeDense:ls,validateMatrix:fs}});var lt=x((Cr,qt)=>{var hs=new Uint32Array([1116352408,1899447441,3049323471,3921009573,961987163,1508970993,2453635748,2870763221,3624381080,310598401,607225278,1426881987,1925078388,2162078206,2614888103,3248222580,3835390401,4022224774,264347078,604807628,770255983,1249150122,1555081692,1996064986,2554220882,2821834349,2952996808,3210313671,3336571891,3584528711,113926993,338241895,666307205,773529912,1294757372,1396182291,1695183700,1986661051,2177026350,2456956037,2730485921,2820302411,3259730800,3345764771,3516065817,3600352804,4094571909,275423344,430227734,506948616,659060556,883997877,958139571,1322822218,1537002063,1747873779,1955562222,2024104815,2227730452,2361852424,2428436474,2756734187,3204031479,3329325298]),N=new Uint32Array(64);function q(s,t){return s>>>t|s<<32-t}function us(s,t,n){for(let l=0;l<16;l++)N[l]=(t[n]<<24|t[n+1]<<16|t[n+2]<<8|t[n+3])>>>0,n+=4;for(let l=16;l<64;l++){let d=q(N[l-15],7)^q(N[l-15],18)^N[l-15]>>>3,u=q(N[l-2],17)^q(N[l-2],19)^N[l-2]>>>10;N[l]=N[l-16]+d+N[l-7]+u>>>0}let e=s[0],o=s[1],r=s[2],i=s[3],a=s[4],f=s[5],h=s[6],c=s[7];for(let l=0;l<64;l++){let d=q(a,6)^q(a,11)^q(a,25),u=a&f^~a&h,p=c+d+u+hs[l]+N[l]>>>0,m=q(e,2)^q(e,13)^q(e,22),w=e&o^e&r^o&r,b=m+w>>>0;c=h,h=f,f=a,a=i+p>>>0,i=r,r=o,o=e,e=p+b>>>0}s[0]=s[0]+e>>>0,s[1]=s[1]+o>>>0,s[2]=s[2]+r>>>0,s[3]=s[3]+i>>>0,s[4]=s[4]+a>>>0,s[5]=s[5]+f>>>0,s[6]=s[6]+h>>>0,s[7]=s[7]+c>>>0}function ds(s){let t=s instanceof Uint8Array?s:new Uint8Array(s),n=t.length,e=n*8,o=n+9+63&-64,r=new Uint8Array(o);r.set(t),r[n]=128;let i=new DataView(r.buffer);i.setUint32(o-4,e>>>0,!1),e>4294967295&&i.setUint32(o-8,e/4294967296>>>0,!1);let a=new Uint32Array([1779033703,3144134277,1013904242,2773480762,1359893119,2600822924,528734635,1541459225]);for(let h=0;h<o;h+=64)us(a,r,h);let f="";for(let h=0;h<8;h++)f+=a[h].toString(16).padStart(8,"0");return f}qt.exports={sha256Sync:ds}});var $t=x((Wr,_t)=>{var ps=new Uint8Array([87,76,82,78]),ms=1,ws=16,gs={FLOAT32:"float32",FLOAT64:"float64",INT32:"int32"};_t.exports={BUNDLE_MAGIC:ps,BUNDLE_VERSION:ms,HEADER_SIZE:ws,DTYPE:gs}});var C=x((Jr,Dt)=>{var{BUNDLE_MAGIC:Ot,BUNDLE_VERSION:G,HEADER_SIZE:I}=$t(),{BundleError:$}=k(),{sha256Sync:Lt}=lt();function T(s){return s==null||typeof s!="object"?JSON.stringify(s):Array.isArray(s)?"["+s.map(n=>T(n)).join(",")+"]":"{"+Object.keys(s).sort().map(n=>JSON.stringify(n)+":"+T(s[n])).join(",")+"}"}var ft=new TextEncoder,ht=new TextDecoder;function bs(s,t){let n=[...t].sort((u,p)=>u.id<p.id?-1:u.id>p.id?1:0),e=0,o=[],r=[];for(let u of n){let p=u.data instanceof Uint8Array?u.data:new Uint8Array(u.data),m=Lt(p),w={id:u.id,offset:e,length:p.length,sha256:m};u.mediaType&&(w.mediaType=u.mediaType),o.push(w),r.push(p),e+=p.length}let i={...s,bundleVersion:G},a=ft.encode(T(i)),f=ft.encode(T(o)),h=I+a.length+f.length+e,c=new Uint8Array(h),l=new DataView(c.buffer);c.set(Ot,0),l.setUint32(4,G,!0),l.setUint32(8,a.length,!0),l.setUint32(12,f.length,!0),c.set(a,I),c.set(f,I+a.length);let d=I+a.length+f.length;for(let u of r)c.set(u,d),d+=u.length;return c}function Bt(s){let t=s instanceof Uint8Array?s:new Uint8Array(s);if(t.length<I)throw new $(`Bundle too small: ${t.length} bytes (minimum ${I})`);for(let l=0;l<4;l++)if(t[l]!==Ot[l])throw new $("Invalid bundle magic (expected WLRN)");let n=new DataView(t.buffer,t.byteOffset,t.byteLength),e=n.getUint32(4,!0);if(e!==G)throw new $(`Unsupported bundle version: ${e} (expected ${G})`);let o=n.getUint32(8,!0),r=n.getUint32(12,!0);if(I+o+r>t.length)throw new $(`Bundle truncated: header declares ${I+o+r} bytes but got ${t.length}`);let i,a;try{i=JSON.parse(ht.decode(t.subarray(I,I+o)))}catch(l){throw new $(`Invalid manifest JSON: ${l.message}`)}try{a=JSON.parse(ht.decode(t.subarray(I+o,I+o+r)))}catch(l){throw new $(`Invalid TOC JSON: ${l.message}`)}let f=I+o+r,h=t.length-f;for(let l=0;l<a.length;l++){let d=a[l];if(d.offset<0||d.length<0||d.offset+d.length>h)throw new $(`TOC entry "${d.id}" out of bounds: offset=${d.offset}, length=${d.length}, blobRegion=${h}`);for(let u=l+1;u<a.length;u++){let p=a[u],m=d.offset,w=d.offset+d.length,b=p.offset,g=p.offset+p.length;if(m<g&&b<w&&d.length>0&&p.length>0)throw new $(`TOC entries "${d.id}" and "${p.id}" overlap`)}}let c=t.subarray(f);return{manifest:i,toc:a,blobs:c}}function ys(s){return ft.encode(T(s))}function xs(s){let t=s instanceof Uint8Array?s:new Uint8Array(s);return JSON.parse(ht.decode(t))}function As(s){let{manifest:t,toc:n,blobs:e}=Bt(s);for(let o of n){let r=e.subarray(o.offset,o.offset+o.length),i=Lt(r);if(i!==o.sha256)throw new $(`SHA-256 mismatch for "${o.id}": expected ${o.sha256}, got ${i}`)}return{manifest:t,toc:n,blobs:e}}Dt.exports={encodeBundle:bs,decodeBundle:Bt,encodeJSON:ys,decodeJSON:xs,validateBundle:As}});var K=x((Gr,Rt)=>{var{RegistryError:D}=k(),{decodeBundle:zt}=C(),V=new Map;function Fs(s,t){if(typeof s!="string"||!s.includes("@"))throw new D(`Invalid typeId "${s}": must contain "@" (e.g. "wlearn.liblinear.classifier@1")`);if(typeof t!="function")throw new D("loaderFn must be a function");V.set(s,t)}async function Ss(s){let{manifest:t,toc:n,blobs:e}=zt(s),{typeId:o}=t;if(!o)throw new D("Bundle manifest missing typeId");let r=V.get(o);if(!r){let i=[...V.keys()],a=i.length>0?`Registered loaders: ${i.join(", ")}`:"No loaders registered";throw new D(`No loader registered for typeId "${o}". ${a}. Install the corresponding @wlearn/* package and import it to register the loader.`)}return await r(t,n,e)}function Es(s){let{manifest:t,toc:n,blobs:e}=zt(s),{typeId:o}=t;if(!o)throw new D("Bundle manifest missing typeId");let r=V.get(o);if(!r){let a=[...V.keys()],f=a.length>0?`Registered loaders: ${a.join(", ")}`:"No loaders registered";throw new D(`No loader registered for typeId "${o}". ${f}. Install the corresponding @wlearn/* package and import it to register the loader.`)}let i=r(t,n,e);if(i&&typeof i.then=="function")throw new D(`Loader for "${o}" returned a Promise. Use async load() instead of loadSync().`);return i}function Is(){return new Map(V)}Rt.exports={register:Fs,load:Ss,loadSync:Es,getRegistry:Is}});var dt=x((Kr,jt)=>{var ut=class{#t;#e;constructor(t,n){this.#t=t,this.#e=n}get name(){return this.#t}get estimator(){return this.#e}get isFitted(){return this.#e.isFitted}get isTransformer(){return typeof this.#e.transform=="function"}};jt.exports={Step:ut}});var Tt=x((Hr,Yt)=>{var{Step:Ps}=dt(),{DisposedError:vs,NotFittedError:Ns,ValidationError:H}=k(),{encodeBundle:ks,decodeBundle:Ms}=C(),{register:qs,load:Ut}=K(),Vt="wlearn.pipeline@1",Xt=!1,Z=class s{#t;#e=!1;#r=!1;constructor(t){if(this.#t=t.map(([n,e])=>new Ps(n,e)),this.#t.length===0)throw new H("Pipeline requires at least one step")}#n(){if(this.#r)throw new vs("Pipeline has been disposed.")}#o(){if(this.#n(),!this.#e)throw new Ns("Pipeline is not fitted. Call fit() first.")}#s(t){let n=t;for(let e=0;e<this.#t.length-1;e++)n=this.#t[e].estimator.transform(n);return n}fit(t,n){this.#n();let e=t;for(let r=0;r<this.#t.length-1;r++){let i=this.#t[r].estimator;typeof i.fitTransform=="function"?e=i.fitTransform(e,n):(i.fit(e,n),e=i.transform(e))}return this.#t[this.#t.length-1].estimator.fit(e,n),this.#e=!0,this}predict(t){this.#o();let n=this.#s(t);return this.#t[this.#t.length-1].estimator.predict(n)}predictProba(t){this.#o();let n=this.#t[this.#t.length-1].estimator;if(typeof n.predictProba!="function")throw new H("Last step does not support predictProba");let e=this.#s(t);return n.predictProba(e)}score(t,n){this.#o();let e=this.#s(t);return this.#t[this.#t.length-1].estimator.score(e,n)}save(){this.#o();let t={typeId:Vt,steps:this.#t.map(e=>({name:e.name,params:e.estimator.getParams()}))},n=this.#t.map(e=>({id:e.name,data:e.estimator.save(),mediaType:"application/x-wlearn-bundle"}));return ks(t,n)}static async load(t){let{manifest:n,toc:e,blobs:o}=Ms(t),r=[];for(let a of n.steps){let f=e.find(l=>l.id===a.name);if(!f)throw new H(`No artifact found for pipeline step "${a.name}"`);let h=o.subarray(f.offset,f.offset+f.length),c=await Ut(h);r.push([a.name,c])}let i=new s(r);return i.#e=!0,i}dispose(){if(!this.#r){this.#r=!0;for(let t of this.#t)t.estimator.dispose()}}getParams(){let t={};for(let n of this.#t)t[n.name]=n.estimator.getParams();return t}setParams(t){this.#n();for(let n of this.#t)t[n.name]&&n.estimator.setParams(t[n.name]);return this}get capabilities(){return this.#t[this.#t.length-1].estimator.capabilities}get isFitted(){return this.#e}static registerLoader(){Xt||(Xt=!0,qs(Vt,(t,n,e)=>s._loadFromParts(t,n,e)))}static async _loadFromParts(t,n,e){let o=[];for(let i of t.steps){let a=n.find(c=>c.id===i.name);if(!a)throw new H(`No artifact found for pipeline step "${i.name}"`);let f=e.subarray(a.offset,a.offset+a.length),h=await Ut(f);o.push([i.name,h])}let r=new s(o);return r.#e=!0,r}};Z.registerLoader();Yt.exports={Pipeline:Z}});var Gt=x((Qr,Jt)=>{var{normalizeX:Ct,normalizeY:Zr}=U(),{ValidationError:Wt}=k(),pt=class s{#t;#e=!1;#r=null;#n=null;#o=null;#s=null;#i=0;constructor(t={}){this.#t={impute:t.impute??"auto",encode:t.encode??"auto",scale:t.scale??!1,maxCategories:t.maxCategories??20,...t}}fit(t,n){let e=Ct(t),{data:o,rows:r,cols:i}=e;this.#r=new Array(i),this.#n=new Float64Array(i);let a=new Array(i);for(let h=0;h<i;h++){let c=[],l=!1,d=0,u=0;for(let b=0;b<r;b++){let g=o[b*i+h];Number.isNaN(g)?l=!0:(c.push(g),d+=g,u++)}let p=new Set(c),m=c.every(b=>b===Math.floor(b)),w=this.#t.encode!==!1&&m&&p.size<=this.#t.maxCategories&&p.size>=2;if(this.#r[h]=w?"categorical":"numeric",this.#t.impute!==!1&&l)if(w){let b=new Map;for(let F of c)b.set(F,(b.get(F)||0)+1);let g=c[0],y=0;for(let[F,E]of b)E>y&&(g=F,y=E);this.#n[h]=g}else this.#n[h]=u>0?d/u:0;a[h]={values:c,unique:p,sum:d,count:u}}this.#o=new Array(i).fill(null);let f=0;if(this.#t.encode!==!1)for(let h=0;h<i;h++)if(this.#r[h]==="categorical"){let c=[...a[h].unique].sort((l,d)=>l-d);if(this.#t.encode==="label"||this.#t.encode===!1){let l=new Map;c.forEach((d,u)=>l.set(d,u)),this.#o[h]={type:"label",mapping:l,startIdx:f},f++}else{let l=new Map;c.forEach((d,u)=>l.set(d,f+u)),this.#o[h]={type:"onehot",mapping:l,startIdx:f,size:c.length},f+=c.length}}else f++;else f=i;if(this.#i=f,this.#s=new Array(this.#i).fill(null),this.#t.scale){let h=0;for(let c=0;c<i;c++){if(this.#r[c]==="categorical"&&this.#t.encode!==!1){let d=this.#o[c];h+=d.type==="onehot"?d.size:1;continue}let l=a[c].values;if(this.#t.scale==="standard"){let d=a[c].sum/(a[c].count||1),u=0;for(let m of l)u+=(m-d)**2;u/=l.length||1;let p=Math.sqrt(u);this.#s[h]={mean:d,std:p||1}}else if(this.#t.scale==="minmax"){let d=1/0,u=-1/0;for(let m of l)m<d&&(d=m),m>u&&(u=m);let p=u-d||1;this.#s[h]={min:d,range:p}}h++}}return this.#e=!0,this}transform(t){if(!this.#e)throw new Wt("Preprocessor not fitted");let n=Ct(t),{data:e,rows:o,cols:r}=n,i=this.#i,a=new Float64Array(o*i);for(let f=0;f<o;f++){let h=0;for(let c=0;c<r;c++){let l=e[f*r+c];if(Number.isNaN(l)&&this.#t.impute!==!1&&(l=this.#n[c]),this.#r[c]==="categorical"&&this.#o[c]){let d=this.#o[c];if(d.type==="onehot"){let u=d.mapping.get(l);u!==void 0&&(a[f*i+u]=1),h=d.startIdx+d.size}else{let u=d.mapping.get(l);a[f*i+h]=u!==void 0?u:-1,h++}}else a[f*i+h]=l,h++}if(this.#t.scale)for(let c=0;c<i;c++){let l=this.#s[c];l&&(this.#t.scale==="standard"?a[f*i+c]=(a[f*i+c]-l.mean)/l.std:this.#t.scale==="minmax"&&(a[f*i+c]=(a[f*i+c]-l.min)/l.range))}}return{data:a,rows:o,cols:i}}fitTransform(t,n){return this.fit(t,n),this.transform(t)}getState(){if(!this.#e)throw new Wt("Preprocessor not fitted");return{config:this.#t,colTypes:this.#r,imputeValues:[...this.#n],encodings:this.#o.map(t=>t?{type:t.type,mapping:[...t.mapping.entries()],startIdx:t.startIdx,size:t.size}:null),scaleParams:this.#s,outputCols:this.#i}}static fromState(t){let n=new s(t.config);return n.#r=t.colTypes,n.#n=new Float64Array(t.imputeValues),n.#o=t.encodings.map(e=>e?{type:e.type,mapping:new Map(e.mapping),startIdx:e.startIdx,size:e.size}:null),n.#s=t.scaleParams,n.#i=t.outputCols,n.#e=!0,n}get isFitted(){return this.#e}get outputCols(){return this.#i}get capabilities(){return{transformer:!0}}dispose(){}};Jt.exports={Preprocessor:pt}});var oe=x((to,re)=>{var{NotFittedError:Kt,DisposedError:Ht,ValidationError:X}=k(),{normalizeX:Q}=U(),{encodeBundle:Zt,encodeJSON:Qt,decodeJSON:te}=C(),{register:ee}=K(),se="wlearn.preprocess.standard_scaler@1",ne="wlearn.preprocess.minmax_scaler@1",tt=class s{#t=null;#e=null;#r=!1;#n=!1;#o={};constructor(t={}){this.#o={...t}}fit(t){this.#s();let{rows:n,cols:e,data:o}=Q(t);if(n===0)throw new X("Cannot fit on empty data");let r=new Float64Array(e),i=new Float64Array(e);for(let f=0;f<n;f++){let h=f+1;for(let c=0;c<e;c++){let l=o[f*e+c],d=l-r[c];r[c]+=d/h;let u=l-r[c];i[c]+=d*u}}let a=new Float64Array(e);for(let f=0;f<e;f++)a[f]=n>1?Math.sqrt(i[f]/(n-1)):0;return this.#t=r,this.#e=a,this.#r=!0,this}transform(t){this.#i();let{rows:n,cols:e,data:o}=Q(t);if(e!==this.#t.length)throw new X(`Expected ${this.#t.length} columns, got ${e}`);let r=new Float64Array(n*e);for(let i=0;i<n;i++)for(let a=0;a<e;a++){let f=i*e+a,h=this.#e[a];r[f]=h>0?(o[f]-this.#t[a])/h:0}return{rows:n,cols:e,data:r}}fitTransform(t){return this.fit(t),this.transform(t)}save(){this.#i();let t={means:Array.from(this.#t),stds:Array.from(this.#e)};return Zt({typeId:se,params:this.getParams()},[{id:"params",data:Qt(t),mediaType:"application/json"}])}static _fromBundle(t,n,e){let o=n.find(a=>a.id==="params");if(!o)throw new X('Bundle missing "params" artifact');let r=te(e.subarray(o.offset,o.offset+o.length)),i=new s(t.params||{});return i.#t=new Float64Array(r.means),i.#e=new Float64Array(r.stds),i.#r=!0,i}dispose(){this.#n||(this.#n=!0,this.#t=null,this.#e=null,this.#r=!1)}getParams(){return{...this.#o}}setParams(t){return Object.assign(this.#o,t),this}get isFitted(){return this.#r&&!this.#n}#s(){if(this.#n)throw new Ht("StandardScaler has been disposed.")}#i(){if(this.#s(),!this.#r)throw new Kt("StandardScaler is not fitted.")}},et=class s{#t=null;#e=null;#r=!1;#n=!1;#o={};constructor(t={}){this.#o={...t}}fit(t){this.#s();let{rows:n,cols:e,data:o}=Q(t);if(n===0)throw new X("Cannot fit on empty data");let r=new Float64Array(e).fill(1/0),i=new Float64Array(e).fill(-1/0);for(let a=0;a<n;a++)for(let f=0;f<e;f++){let h=o[a*e+f];h<r[f]&&(r[f]=h),h>i[f]&&(i[f]=h)}return this.#t=r,this.#e=i,this.#r=!0,this}transform(t){this.#i();let{rows:n,cols:e,data:o}=Q(t);if(e!==this.#t.length)throw new X(`Expected ${this.#t.length} columns, got ${e}`);let r=new Float64Array(n*e);for(let i=0;i<n;i++)for(let a=0;a<e;a++){let f=i*e+a,h=this.#e[a]-this.#t[a];r[f]=h>0?(o[f]-this.#t[a])/h:0}return{rows:n,cols:e,data:r}}fitTransform(t){return this.fit(t),this.transform(t)}save(){this.#i();let t={mins:Array.from(this.#t),maxs:Array.from(this.#e)};return Zt({typeId:ne,params:this.getParams()},[{id:"params",data:Qt(t),mediaType:"application/json"}])}static _fromBundle(t,n,e){let o=n.find(a=>a.id==="params");if(!o)throw new X('Bundle missing "params" artifact');let r=te(e.subarray(o.offset,o.offset+o.length)),i=new s(t.params||{});return i.#t=new Float64Array(r.mins),i.#e=new Float64Array(r.maxs),i.#r=!0,i}dispose(){this.#n||(this.#n=!0,this.#t=null,this.#e=null,this.#r=!1)}getParams(){return{...this.#o}}setParams(t){return Object.assign(this.#o,t),this}get isFitted(){return this.#r&&!this.#n}#s(){if(this.#n)throw new Ht("MinMaxScaler has been disposed.")}#i(){if(this.#s(),!this.#r)throw new Kt("MinMaxScaler is not fitted.")}};ee(se,tt._fromBundle);ee(ne,et._fromBundle);re.exports={StandardScaler:tt,MinMaxScaler:et}});var mt=x((eo,ie)=>{function _s(s=42){let t=s|0;return()=>(t=t*1664525+1013904223&2147483647,t/2147483647)}function $s(s,t){for(let n=s.length-1;n>0;n--){let e=t()*(n+1)|0,o=s[n];s[n]=s[e],s[e]=o}return s}ie.exports={makeLCG:_s,shuffle:$s}});var bt=x((so,ae)=>{var{ValidationError:L}=k();function B(s,t,n){if(!s||!t||s.length===0||t.length===0)throw new L(`${n}: inputs must be non-empty`);if(s.length!==t.length)throw new L(`${n}: length mismatch (${s.length} vs ${t.length})`)}function Y(s,t){let n=new Set;for(let r=0;r<s.length;r++)n.add(s[r]);if(t)for(let r=0;r<t.length;r++)n.add(t[r]);let e=new Int32Array([...n].sort((r,i)=>r-i)),o=new Map;for(let r=0;r<e.length;r++)o.set(e[r],r);return{labels:e,labelMap:o,nClasses:e.length}}function st(s,t,n,e){let o=new Int32Array(e*e);for(let r=0;r<s.length;r++){let i=n.get(s[r]),a=n.get(t[r]);o[i*e+a]++}return o}function wt(s,t){let n=new Float64Array(t),e=new Float64Array(t),o=new Float64Array(t);for(let r=0;r<t;r++){n[r]=s[r*t+r];for(let i=0;i<t;i++)i!==r&&(e[r]+=s[i*t+r],o[r]+=s[r*t+i])}return{tp:n,fp:e,fn:o}}function Os(s,t){B(s,t,"accuracy");let n=0;for(let e=0;e<s.length;e++)s[e]===t[e]&&n++;return n/s.length}function Ls(s,t){B(s,t,"r2Score");let n=s.length,e=0;for(let i=0;i<n;i++)e+=s[i];e/=n;let o=0,r=0;for(let i=0;i<n;i++){let a=s[i]-e;o+=a*a;let f=s[i]-t[i];r+=f*f}return o===0?0:1-r/o}function Bs(s,t){B(s,t,"meanSquaredError");let n=0;for(let e=0;e<s.length;e++){let o=s[e]-t[e];n+=o*o}return n/s.length}function Ds(s,t){B(s,t,"meanAbsoluteError");let n=0;for(let e=0;e<s.length;e++)n+=Math.abs(s[e]-t[e]);return n/s.length}function zs(s,t){B(s,t,"confusionMatrix");let{labels:n,labelMap:e,nClasses:o}=Y(s,t);return{matrix:st(s,t,e,o),labels:n}}function gt(s,t,n){if(s==="binary"){if(t>2)throw new L('average="binary" requires exactly 2 classes');return"binary"}return s||"binary"}function Rs(s,t,{average:n="binary"}={}){B(s,t,"precisionScore");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fp:f}=wt(i,r),h=gt(n,r,e);if(h==="binary"){let l=r-1,d=a[l]+f[l];return d===0?0:a[l]/d}if(h==="micro"){let l=0,d=0;for(let u=0;u<r;u++)l+=a[u],d+=f[u];return l+d===0?0:l/(l+d)}let c=0;for(let l=0;l<r;l++){let d=a[l]+f[l];c+=d===0?0:a[l]/d}return c/r}function js(s,t,{average:n="binary"}={}){B(s,t,"recallScore");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fn:f}=wt(i,r),h=gt(n,r,e);if(h==="binary"){let l=r-1,d=a[l]+f[l];return d===0?0:a[l]/d}if(h==="micro"){let l=0,d=0;for(let u=0;u<r;u++)l+=a[u],d+=f[u];return l+d===0?0:l/(l+d)}let c=0;for(let l=0;l<r;l++){let d=a[l]+f[l];c+=d===0?0:a[l]/d}return c/r}function Us(s,t,{average:n="binary"}={}){B(s,t,"f1Score");let{labels:e,labelMap:o,nClasses:r}=Y(s,t),i=st(s,t,o,r),{tp:a,fp:f,fn:h}=wt(i,r),c=gt(n,r,e);function l(u,p,m){let w=u+p===0?0:u/(u+p),b=u+m===0?0:u/(u+m);return w+b===0?0:2*w*b/(w+b)}if(c==="binary"){let u=r-1;return l(a[u],f[u],h[u])}if(c==="micro"){let u=0,p=0,m=0;for(let w=0;w<r;w++)u+=a[w],p+=f[w],m+=h[w];return l(u,p,m)}let d=0;for(let u=0;u<r;u++)d+=l(a[u],f[u],h[u]);return d/r}function Vs(s,t,{nClasses:n,eps:e=1e-15}={}){if(!s||s.length===0)throw new L("logLoss: yTrue must be non-empty");let o=s.length;if(!n){let a=new Set;for(let f=0;f<o;f++)a.add(s[f]);n=a.size}if(t.length!==o*n)throw new L(`logLoss: yProba length (${t.length}) must be n * nClasses (${o*n})`);let{labelMap:r}=Y(s),i=0;for(let a=0;a<o;a++){let f=r.get(s[a]),h=t[a*n+f];h=Math.max(e,Math.min(1-e,h)),i-=Math.log(h)}return i/o}function Xs(s,t){if(!s||s.length===0)throw new L("rocAuc: yTrue must be non-empty");if(s.length!==t.length)throw new L("rocAuc: length mismatch");let{labels:n}=Y(s);if(n.length!==2)throw new L("rocAuc: requires exactly 2 classes");let e=n[n.length-1],o=s.length,r=Array.from({length:o},(u,p)=>p);r.sort((u,p)=>t[p]-t[u]);let i=0,a=0;for(let u=0;u<o;u++)s[u]===e?i++:a++;if(i===0||a===0)return 0;let f=0,h=0,c=0,l=0,d=0;for(let u=0;u<o;u++){let p=r[u];s[p]===e?f++:h++;let m=f/i,w=h/a;d+=(w-l)*(m+c)/2,c=m,l=w}return d}ae.exports={accuracy:Os,r2Score:Ls,meanSquaredError:Bs,meanAbsoluteError:Ds,confusionMatrix:zs,precisionScore:Rs,recallScore:js,f1Score:Us,logLoss:Vs,rocAuc:Xs}});var pe=x((no,de)=>{var{ValidationError:z}=k(),{makeLCG:yt,shuffle:xt}=mt(),{normalizeX:Ys,normalizeY:Ts}=U(),{accuracy:Cs,r2Score:Ws,meanSquaredError:Js,meanAbsoluteError:Gs}=bt(),ce={accuracy:(s,t)=>Cs(s,t),r2:(s,t)=>Ws(s,t),neg_mse:(s,t)=>-Js(s,t),neg_mae:(s,t)=>-Gs(s,t)};function he(s){if(typeof s=="function")return s;let t=ce[s];if(!t)throw new z(`Unknown scoring: "${s}". Available: ${Object.keys(ce).join(", ")}`);return t}function Ks(s,t=5,{shuffle:n=!0,seed:e=42}={}){if(s<t)throw new z(`kFold: n (${s}) must be >= k (${t})`);if(t<2)throw new z("kFold: k must be >= 2");let o=Int32Array.from({length:s},(h,c)=>c);if(n){let h=yt(e);xt(o,h)}let r=Math.floor(s/t),i=s%t,a=[],f=0;for(let h=0;h<t;h++){let c=r+(h<i?1:0),l=o.slice(f,f+c),d=[];f>0&&d.push(o.slice(0,f)),f+c<s&&d.push(o.slice(f+c));let u=Qs(d);a.push({train:u,test:l}),f+=c}return a}function ue(s,t=5,{shuffle:n=!0,seed:e=42}={}){let o=s.length;if(o<t)throw new z(`stratifiedKFold: n (${o}) must be >= k (${t})`);if(t<2)throw new z("stratifiedKFold: k must be >= 2");let r=new Map;for(let h=0;h<o;h++){let c=s[h];r.has(c)||r.set(c,[]),r.get(c).push(h)}if(n){let h=yt(e);for(let c of r.values())xt(c,h)}let i=Array.from({length:t},()=>[]);for(let h of r.values())for(let c=0;c<h.length;c++)i[c%t].push(h[c]);let a=Int32Array.from({length:o},(h,c)=>c),f=[];for(let h=0;h<t;h++){let c=new Set(i[h]),l=new Int32Array(i[h]),d=a.filter(u=>!c.has(u));f.push({train:d,test:l})}return f}function Hs(s,{testSize:t=.2,shuffle:n=!0,seed:e=42}={}){if(s<2)throw new z("trainTestSplit: n must be >= 2");let o=Math.max(1,Math.round(s*t)),r=s-o;if(r<1)throw new z("trainTestSplit: testSize too large");let i=Int32Array.from({length:s},(a,f)=>f);if(n){let a=yt(e);xt(i,a)}return{train:i.slice(0,r),test:i.slice(r)}}async function Zs(s,t,n,{cv:e=5,scoring:o="accuracy",seed:r=42,params:i={}}={}){let a=Ys(t),f=Ts(n),h=he(o),c;Array.isArray(e)?c=e:c=ue(f,e,{shuffle:!0,seed:r});let l=new Float64Array(c.length);for(let d=0;d<c.length;d++){let{train:u,test:p}=c[d],m=le(a,u),w=fe(f,u),b=le(a,p),g=fe(f,p),y=await s.create(i);try{y.fit(m,w);let F=await y.predict(b);l[d]=h(g,F)}finally{y.dispose()}}return l}function Qs(s){let t=0;for(let o of s)t+=o.length;let n=new Int32Array(t),e=0;for(let o of s)n.set(o,e),e+=o.length;return n}function le(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function fe(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}de.exports={kFold:Ks,stratifiedKFold:ue,trainTestSplit:Hs,crossValScore:Zs,getScorer:he}});var ge=x((ro,we)=>{function me(s){return s!=null&&typeof s.then=="function"}function tn(s,t){return me(s)?s.then(t):t(s)}we.exports={isPromiseLike:me,lift:tn}});var Ae=x((oo,xe)=>{var{normalizeY:en}=U(),{round:sn}=Math;function be(s){let t=en(s);if(t instanceof Int32Array)return"classification";let n=new Set;for(let e=0;e<t.length;e++){if(t[e]!==sn(t[e]))return"regression";n.add(t[e])}return n.size<=20?"classification":"regression"}var ye=new WeakMap;function P(s){let t=ye.get(s);if(!t)throw new Error("Model: invalid instance");return t}function W(s,t){let n=P(s);if(n.disposed)throw new Error(`${t} has been disposed.`);if(!n.inner)throw new Error(`${t}: not fitted`);return n.inner}function nn(s,t,n={}){let e=n.name||"Model",o=n.load||null,r=s===t;class i{constructor(c,l){ye.set(this,{inner:null,task:c,params:l,disposed:!1})}static async create(c={}){let l={...c},d=l.task||null;delete l.task,o&&await o();let u=new i(d,l),p=P(u);if(d){let m=d==="classification"?s:t;p.inner=await m.create({...l,task:d})}else r&&(p.inner=await s.create(l));return u}static async load(c){try{let l=await s.load(c),d=new i("classification",{}),u=P(d);return u.inner=l,typeof l.getParams=="function"&&(u.params=l.getParams(),u.params.task&&(u.task=u.params.task)),d}catch{let d=await t.load(c),u=new i("regression",{}),p=P(u);return p.inner=d,typeof d.getParams=="function"&&(p.params=d.getParams(),p.params.task&&(p.task=p.params.task)),u}}fit(c,l,d){let u=P(this);if(u.task||(u.task=be(l)),!u.inner){let p=u.task==="classification"?s:t;u.inner=new p({...u.params,task:u.task})}return u.inner.fit(c,l,d),this}predict(c,l){return W(this,e).predict(c,l)}predictProba(c,l){let d=W(this,e);if(typeof d.predictProba!="function")throw new Error(`${e}: predictProba not available`);return d.predictProba(c,l)}score(c,l,d){return W(this,e).score(c,l,d)}save(){return W(this,e).save()}dispose(){let c=P(this);c.inner&&typeof c.inner.dispose=="function"&&c.inner.dispose(),c.inner=null,c.disposed=!0}getParams(){let c=P(this),l=c.inner&&typeof c.inner.getParams=="function"?c.inner.getParams():{...c.params};return c.task&&(l.task=c.task),l}setParams(c){let l=P(this),d={...c};if(d.task){let u=d.task;delete d.task,u!==l.task&&l.inner&&this.dispose(),l.task=u}return l.params=d,l.inner&&typeof l.inner.setParams=="function"&&l.inner.setParams(d),this}get task(){return P(this).task}get isFitted(){let c=P(this);return c.inner?c.inner.isFitted!==void 0?c.inner.isFitted:!0:!1}get classes(){let c=P(this);return c.inner?typeof c.inner.classes=="function"?c.inner.classes():c.inner.classes!==void 0?c.inner.classes:new Int32Array(0):new Int32Array(0)}get capabilities(){let c=P(this);return c.inner&&c.inner.capabilities!==void 0?c.inner.capabilities:{}}}let a=new Set(["constructor","fit","predict","predictProba","score","save","dispose","getParams","setParams"]),f=new Set(["classes","task","isFitted","capabilities"]);for(let h of[s,t])if(!(!h||!h.prototype))for(let c of Object.getOwnPropertyNames(h.prototype)){if(c.startsWith("_")||c.startsWith("#")||a.has(c)||f.has(c)||c in i.prototype)continue;let l=Object.getOwnPropertyDescriptor(h.prototype,c);l&&(typeof l.value=="function"?i.prototype[c]=function(...d){return W(this,e)[c](...d)}:l.get&&Object.defineProperty(i.prototype,c,{get(){let d=P(this);if(d.inner)return d.inner[c]},configurable:!0}))}return(s.defaultSearchSpace||t.defaultSearchSpace)&&(i.defaultSearchSpace=()=>s.defaultSearchSpace?.()||t.defaultSearchSpace?.()||{}),(s.budgetSpec||t.budgetSpec)&&(i.budgetSpec=()=>s.budgetSpec?.()||t.budgetSpec?.()||void 0),Object.defineProperty(i,"name",{value:e,configurable:!0}),i}xe.exports={createModelClass:nn,detectTask:be}});var R=x((io,Fe)=>{var{WlearnError:rn,BundleError:on,RegistryError:an,ValidationError:cn,NotFittedError:ln,DisposedError:fn}=k(),{normalizeX:hn,normalizeY:un,makeDense:dn,validateMatrix:pn}=U(),{sha256Sync:mn}=lt(),{encodeBundle:wn,decodeBundle:gn,validateBundle:bn,encodeJSON:yn,decodeJSON:xn}=C(),{register:An,load:Fn,loadSync:Sn,getRegistry:En}=K(),{Pipeline:In}=Tt(),{Step:Pn}=dt(),{Preprocessor:vn}=Gt(),{StandardScaler:Nn,MinMaxScaler:kn}=oe(),{makeLCG:Mn,shuffle:qn}=mt(),{accuracy:_n,r2Score:$n,meanSquaredError:On,meanAbsoluteError:Ln,confusionMatrix:Bn,precisionScore:Dn,recallScore:zn,f1Score:Rn,logLoss:jn,rocAuc:Un}=bt(),{kFold:Vn,stratifiedKFold:Xn,trainTestSplit:Yn,crossValScore:Tn,getScorer:Cn}=pe(),{isPromiseLike:Wn,lift:Jn}=ge(),{createModelClass:Gn,detectTask:Kn}=Ae();Fe.exports={WlearnError:rn,BundleError:on,RegistryError:an,ValidationError:cn,NotFittedError:ln,DisposedError:fn,normalizeX:hn,normalizeY:un,makeDense:dn,validateMatrix:pn,sha256Sync:mn,encodeBundle:wn,decodeBundle:gn,validateBundle:bn,encodeJSON:yn,decodeJSON:xn,register:An,load:Fn,loadSync:Sn,getRegistry:En,Pipeline:In,Step:Pn,Preprocessor:vn,StandardScaler:Nn,MinMaxScaler:kn,makeLCG:Mn,shuffle:qn,accuracy:_n,r2Score:$n,meanSquaredError:On,meanAbsoluteError:Ln,confusionMatrix:Bn,precisionScore:Dn,recallScore:zn,f1Score:Rn,logLoss:jn,rocAuc:Un,kFold:Vn,stratifiedKFold:Xn,trainTestSplit:Yn,crossValScore:Tn,getScorer:Cn,isPromiseLike:Wn,lift:Jn,createModelClass:Gn,detectTask:Kn}});var qe=x((ao,Me)=>{var{encodeBundle:Hn,decodeBundle:Zn,register:Se,load:Ee,normalizeX:At,normalizeY:Ie,accuracy:Qn,r2Score:tr,ValidationError:nt,NotFittedError:er,DisposedError:sr,lift:Pe}=R(),ve="wlearn.ensemble.voting.classifier@1",Ne="wlearn.ensemble.voting.regressor@1",ke=!1,Ft=class s{#t;#e;#r;#n;#o;#s;#i=!1;#f=!1;constructor(t){this.#t=t.estimators||[],this.#r=t.weights||null,this.#n=t.voting||"soft",this.#o=t.task||"classification",this.#e=null,this.#s=null,s._register()}static async create(t={}){return new s(t)}#l(){if(this.#f)throw new sr("VotingEnsemble has been disposed.")}#a(){if(this.#l(),!this.#i)throw new er("VotingEnsemble is not fitted. Call fit() first.")}async fit(t,n){this.#l();let e=At(t),o=Ie(n);if(this.#o==="classification"){let r=new Set;for(let i=0;i<o.length;i++)r.add(o[i]);this.#s=new Int32Array([...r].sort((i,a)=>i-a))}this.#r||(this.#r=new Float64Array(this.#t.length).fill(1/this.#t.length)),this.#e=[];for(let[r,i,a]of this.#t){let f=await i.create(a||{});f.fit(e,o),this.#e.push(f)}return this.#i=!0,this}predict(t){this.#a();let n=At(t),e=n.rows;if(this.#o==="regression")return this.#c(n,e);if(this.#n==="soft"){let o=this.predictProba(n);return Pe(o,r=>{let i=this.#s.length,a=new Float64Array(e);for(let f=0;f<e;f++){let h=0,c=-1/0;for(let l=0;l<i;l++)r[f*i+l]>c&&(c=r[f*i+l],h=l);a[f]=this.#s[h]}return a})}return this.#h(n,e)}predictProba(t){if(this.#a(),this.#o!=="classification")throw new nt("predictProba is only available for classification");if(this.#n==="hard")throw new nt('predictProba requires voting="soft"');let n=At(t),e=n.rows,o=this.#s.length,r=[],i=!1;for(let f=0;f<this.#e.length;f++){let h=this.#e[f].predictProba(n);h!=null&&typeof h.then=="function"&&(i=!0),r.push(h)}let a=f=>{let h=new Float64Array(e*o);for(let c=0;c<f.length;c++){let l=f[c],d=this.#r[c];for(let u=0;u<e*o;u++)h[u]+=d*l[u]}return h};return i?Promise.all(r).then(a):a(r)}score(t,n){this.#a();let e=this.predict(t),o=Ie(n),r=this.#o==="classification"?Qn:tr;return Pe(e,i=>r(o,i))}save(){this.#a();let n={typeId:this.#o==="classification"?ve:Ne,params:{task:this.#o,voting:this.#n,weights:[...this.#r],estimatorNames:this.#t.map(o=>o[0]),classes:this.#s?[...this.#s]:null}},e=this.#e.map((o,r)=>({id:this.#t[r][0],data:o.save(),mediaType:"application/x-wlearn-bundle"}));return Hn(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=Zn(t),r=n.params,i=new s({task:r.task,voting:r.voting,weights:new Float64Array(r.weights)});i.#s=r.classes?new Int32Array(r.classes):null,i.#t=r.estimatorNames.map(a=>[a,null,null]),i.#e=[];for(let a of r.estimatorNames){let f=e.find(l=>l.id===a);if(!f)throw new nt(`No artifact for estimator "${a}"`);let h=o.subarray(f.offset,f.offset+f.length),c=await Ee(h);i.#e.push(c)}return i.#i=!0,i}dispose(){if(!this.#f&&(this.#f=!0,this.#e))for(let t of this.#e)t.dispose()}getParams(){return{task:this.#o,voting:this.#n,weights:this.#r?[...this.#r]:null,estimatorNames:this.#t.map(t=>t[0])}}setParams(t){return this.#l(),t.voting!==void 0&&(this.#n=t.voting),t.weights!==void 0&&(this.#r=new Float64Array(t.weights)),this}get capabilities(){return{classifier:this.#o==="classification",regressor:this.#o==="regression",predictProba:this.#o==="classification"&&this.#n==="soft",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#i}get classes(){return this.#s}#c(t,n){let e=[],o=!1;for(let i=0;i<this.#e.length;i++){let a=this.#e[i].predict(t);a!=null&&typeof a.then=="function"&&(o=!0),e.push(a)}let r=i=>{let a=new Float64Array(n);for(let f=0;f<i.length;f++){let h=this.#r[f];for(let c=0;c<n;c++)a[c]+=h*i[f][c]}return a};return o?Promise.all(e).then(r):r(e)}#h(t,n){let e=[],o=!1;for(let i=0;i<this.#e.length;i++){let a=this.#e[i].predict(t);a!=null&&typeof a.then=="function"&&(o=!0),e.push(a)}let r=i=>{let a=this.#s.length,f=new Float64Array(n);for(let h=0;h<n;h++){let c=new Float64Array(a);for(let u=0;u<i.length;u++){let p=i[u][h],m=this.#s.indexOf(p);m>=0&&(c[m]+=this.#r[u])}let l=0,d=-1/0;for(let u=0;u<a;u++)c[u]>d&&(d=c[u],l=u);f[h]=this.#s[l]}return f};return o?Promise.all(e).then(r):r(e)}static _register(){if(ke)return;ke=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);Se(ve,t),Se(Ne,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,voting:o.voting,weights:new Float64Array(o.weights)});r.#s=o.classes?new Int32Array(o.classes):null,r.#t=o.estimatorNames.map(i=>[i,null,null]),r.#e=[];for(let i of o.estimatorNames){let a=n.find(c=>c.id===i);if(!a)throw new nt(`No artifact for estimator "${i}"`);let f=e.subarray(a.offset,a.offset+a.length),h=await Ee(f);r.#e.push(h)}return r.#i=!0,r}};Me.exports={VotingEnsemble:Ft}});var Ue=x((co,je)=>{var{encodeBundle:nr,decodeBundle:rr,register:_e,load:$e,normalizeX:Oe,normalizeY:Le,accuracy:or,r2Score:ir,stratifiedKFold:ar,kFold:cr,ValidationError:J,NotFittedError:lr,DisposedError:fr,lift:St}=R(),Be="wlearn.ensemble.stacking.classifier@1",De="wlearn.ensemble.stacking.regressor@1",ze=!1,Et=class s{#t;#e;#r;#n;#o;#s;#i;#f;#l;#a;#c;#h=!1;#d=!1;constructor(t){this.#t=t.estimators||[],this.#e=t.finalEstimator||null,this.#o=t.cv||5,this.#s=t.task||"classification",this.#i=t.passthrough||!1,this.#f=t.seed??42,this.#r=null,this.#n=null,this.#l=null,this.#a=0,this.#c=0,s._register()}static async create(t={}){return new s(t)}#p(){if(this.#d)throw new fr("StackingEnsemble has been disposed.")}#u(){if(this.#p(),!this.#h)throw new lr("StackingEnsemble is not fitted. Call fit() first.")}async fit(t,n){if(this.#p(),!this.#e)throw new J("StackingEnsemble requires a finalEstimator");let e=Oe(t),o=Le(n),r=e.rows;if(this.#s==="classification"){let p=new Set;for(let m=0;m<o.length;m++)p.add(o[m]);this.#l=new Int32Array([...p].sort((m,w)=>m-w)),this.#a=this.#l.length}let i=this.#s==="classification"?ar(o,this.#o,{shuffle:!0,seed:this.#f}):cr(r,this.#o,{shuffle:!0,seed:this.#f}),a=this.#t.length,f=this.#s==="classification"?this.#a:1,h=a*f,c=new Float64Array(r*h);for(let p=0;p<a;p++){let[,m,w]=this.#t[p];for(let{train:b,test:g}of i){let y=Re(e,b),F=hr(o,b),E=Re(e,g),v=await m.create(w||{});try{if(v.fit(y,F),this.#s==="classification"){let S=await v.predictProba(E);for(let A=0;A<g.length;A++){let j=g[A];for(let _=0;_<this.#a;_++)c[j*h+p*f+_]=S[A*this.#a+_]}}else{let S=await v.predict(E);for(let A=0;A<g.length;A++)c[g[A]*h+p]=S[A]}}finally{v.dispose()}}}let l;if(this.#i){this.#c=h+e.cols;let p=new Float64Array(r*this.#c);for(let m=0;m<r;m++)p.set(c.subarray(m*h,(m+1)*h),m*this.#c),p.set(e.data.subarray(m*e.cols,(m+1)*e.cols),m*this.#c+h);l={data:p,rows:r,cols:this.#c}}else this.#c=h,l={data:c,rows:r,cols:h};this.#r=[];for(let[,p,m]of this.#t){let w=await p.create(m||{});w.fit(e,o),this.#r.push(w)}let[,d,u]=this.#e;return this.#n=await d.create(u||{}),this.#n.fit(l,o),this.#h=!0,this}predict(t){this.#u();let n=this.#m(t);return St(n,e=>this.#n.predict(e))}predictProba(t){if(this.#u(),this.#s!=="classification")throw new J("predictProba is only available for classification");if(typeof this.#n.predictProba!="function")throw new J("Meta-model does not support predictProba");let n=this.#m(t);return St(n,e=>this.#n.predictProba(e))}score(t,n){this.#u();let e=this.predict(t),o=Le(n),r=this.#s==="classification"?or:ir;return St(e,i=>r(o,i))}save(){this.#u();let n={typeId:this.#s==="classification"?Be:De,params:{task:this.#s,cv:this.#o,passthrough:this.#i,seed:this.#f,estimatorNames:this.#t.map(o=>o[0]),metaName:this.#e[0],classes:this.#l?[...this.#l]:null,nMetaCols:this.#c}},e=this.#r.map((o,r)=>({id:this.#t[r][0],data:o.save(),mediaType:"application/x-wlearn-bundle"}));return e.push({id:this.#e[0],data:this.#n.save(),mediaType:"application/x-wlearn-bundle"}),nr(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=rr(t);return s._loadFromParts(n,e,o)}dispose(){if(!this.#d){if(this.#d=!0,this.#r)for(let t of this.#r)t.dispose();this.#n&&this.#n.dispose()}}getParams(){return{task:this.#s,cv:this.#o,passthrough:this.#i,seed:this.#f,estimatorNames:this.#t.map(t=>t[0]),metaName:this.#e?this.#e[0]:null}}setParams(t){return this.#p(),t.cv!==void 0&&(this.#o=t.cv),t.passthrough!==void 0&&(this.#i=t.passthrough),t.seed!==void 0&&(this.#f=t.seed),this}get capabilities(){return{classifier:this.#s==="classification",regressor:this.#s==="regression",predictProba:this.#s==="classification",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#h}get classes(){return this.#l}#m(t){let n=Oe(t),e=n.rows,o=this.#r.length,r=this.#s==="classification"?this.#a:1,i=o*r,a=[],f=!1;for(let c=0;c<o;c++){let l=this.#s==="classification"?this.#r[c].predictProba(n):this.#r[c].predict(n);l!=null&&typeof l.then=="function"&&(f=!0),a.push(l)}let h=c=>{let l=new Float64Array(e*this.#c);for(let d=0;d<o;d++)if(this.#s==="classification"){let u=c[d];for(let p=0;p<e;p++)for(let m=0;m<this.#a;m++)l[p*this.#c+d*r+m]=u[p*this.#a+m]}else{let u=c[d];for(let p=0;p<e;p++)l[p*this.#c+d]=u[p]}if(this.#i)for(let d=0;d<e;d++)l.set(n.data.subarray(d*n.cols,(d+1)*n.cols),d*this.#c+i);return{data:l,rows:e,cols:this.#c}};return f?Promise.all(a).then(h):h(a)}static _register(){if(ze)return;ze=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);_e(Be,t),_e(De,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,cv:o.cv,passthrough:o.passthrough,seed:o.seed});r.#l=o.classes?new Int32Array(o.classes):null,r.#a=r.#l?r.#l.length:0,r.#c=o.nMetaCols,r.#t=o.estimatorNames.map(f=>[f,null,null]),r.#e=[o.metaName,null,null],r.#r=[];for(let f of o.estimatorNames){let h=n.find(l=>l.id===f);if(!h)throw new J(`No artifact for base estimator "${f}"`);let c=e.subarray(h.offset,h.offset+h.length);r.#r.push(await $e(c))}let i=n.find(f=>f.id===o.metaName);if(!i)throw new J(`No artifact for meta estimator "${o.metaName}"`);let a=e.subarray(i.offset,i.offset+i.length);return r.#n=await $e(a),r.#h=!0,r}};function Re(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function hr(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}je.exports={StackingEnsemble:Et}});var He=x((lo,Ke)=>{var{encodeBundle:ur,decodeBundle:dr,register:Ve,load:pr,normalizeX:It,normalizeY:Xe,accuracy:mr,r2Score:wr,stratifiedKFold:gr,kFold:br,ValidationError:Ye,NotFittedError:yr,DisposedError:xr,lift:Te}=R(),Ce="wlearn.ensemble.bagged.classifier@1",We="wlearn.ensemble.bagged.regressor@1",Je=!1,Pt=class s{#t;#e;#r;#n;#o;#s;#i;#f=0;#l=0;#a;#c;#h=!1;#d=!1;constructor(t={}){this.#t=t.estimator||null,this.#e=t.kFold||5,this.#r=t.nRepeats||1,this.#n=t.task||"classification",this.#o=t.seed??42,this.#s=null,this.#i=null,this.#a=null,this.#c=null,s._register()}static async create(t={}){return new s(t)}#p(){if(this.#d)throw new xr("BaggedEstimator has been disposed.")}#u(){if(this.#p(),!this.#h)throw new yr("BaggedEstimator is not fitted. Call fit() first.")}async fit(t,n){this.#p();let e=It(t),o=Xe(n),r=e.rows;if(this.#l=r,this.#n==="classification"){let f=new Set;for(let h=0;h<o.length;h++)f.add(o[h]);this.#i=new Int32Array([...f].sort((h,c)=>h-c)),this.#f=this.#i.length}this.#n==="classification"?this.#a=new Float64Array(r*this.#f):this.#a=new Float64Array(r),this.#c=new Uint8Array(r);let[,i,a]=this.#t;this.#s=[];for(let f=0;f<this.#r;f++){let h=this.#o+f,c=this.#n==="classification"?gr(o,this.#e,{shuffle:!0,seed:h}):br(r,this.#e,{shuffle:!0,seed:h});for(let{train:l,test:d}of c){let u=Ge(e,l),p=Ar(o,l),m=Ge(e,d),w=await i.create(a||{});if(w.fit(u,p),this.#n==="classification"){let b=await w.predictProba(m),g=this.#f;for(let y=0;y<d.length;y++){let F=d[y];for(let E=0;E<g;E++)this.#a[F*g+E]+=b[y*g+E]}}else{let b=await w.predict(m);for(let g=0;g<d.length;g++)this.#a[d[g]]+=b[g]}for(let b=0;b<d.length;b++)this.#c[d[b]]+=1;this.#s.push(w)}}return this.#h=!0,this}predict(t){this.#u();let n=It(t),e=n.rows;if(this.#n==="regression")return this.#m(n,e);let o=this.predictProba(n);return Te(o,r=>{let i=this.#f,a=new Float64Array(e);for(let f=0;f<e;f++){let h=0,c=-1/0;for(let l=0;l<i;l++)r[f*i+l]>c&&(c=r[f*i+l],h=l);a[f]=this.#i[h]}return a})}predictProba(t){if(this.#u(),this.#n!=="classification")throw new Ye("predictProba is only available for classification");let n=It(t),e=n.rows,o=this.#f,r=this.#s.length,i=[],a=!1;for(let h of this.#s){let c=h.predictProba(n);c!=null&&typeof c.then=="function"&&(a=!0),i.push(c)}let f=h=>{let c=new Float64Array(e*o);for(let l of h)for(let d=0;d<e*o;d++)c[d]+=l[d];for(let l=0;l<e*o;l++)c[l]/=r;return c};return a?Promise.all(i).then(f):f(i)}score(t,n){this.#u();let e=this.predict(t),o=Xe(n),r=this.#n==="classification"?mr:wr;return Te(e,i=>r(o,i))}#m(t,n){let e=this.#s.length,o=[],r=!1;for(let a of this.#s){let f=a.predict(t);f!=null&&typeof f.then=="function"&&(r=!0),o.push(f)}let i=a=>{let f=new Float64Array(n);for(let h of a)for(let c=0;c<n;c++)f[c]+=h[c];for(let h=0;h<n;h++)f[h]/=e;return f};return r?Promise.all(o).then(i):i(o)}get oofPredictions(){this.#u();let t=new Uint8Array(this.#c);for(let e=0;e<t.length;e++)t[e]===0&&(t[e]=1);if(this.#n==="classification"){let e=this.#f,o=new Float64Array(this.#a);for(let r=0;r<this.#l;r++){let i=t[r];for(let a=0;a<e;a++)o[r*e+a]/=i}return o}let n=new Float64Array(this.#a);for(let e=0;e<this.#l;e++)n[e]/=t[e];return n}save(){this.#u();let n={typeId:this.#n==="classification"?Ce:We,params:{task:this.#n,kFold:this.#e,nRepeats:this.#r,seed:this.#o,estimatorName:this.#t[0],classes:this.#i?[...this.#i]:null,nClasses:this.#f,nSamples:this.#l}},e=this.#s.map((i,a)=>({id:`fold_${a}`,data:i.save(),mediaType:"application/x-wlearn-bundle"})),o=this.oofPredictions,r=new Uint8Array(o.buffer,o.byteOffset,o.byteLength);return e.push({id:"oof",data:r,mediaType:"application/octet-stream"}),ur(n,e)}static async load(t){let{manifest:n,toc:e,blobs:o}=dr(t);return s._loadFromParts(n,e,o)}dispose(){if(!this.#d){if(this.#d=!0,this.#s)for(let t of this.#s)t.dispose();this.#s=null,this.#a=null,this.#c=null}}getParams(){return{task:this.#n,kFold:this.#e,nRepeats:this.#r,seed:this.#o,estimatorName:this.#t?this.#t[0]:null}}setParams(t){return this.#p(),t.kFold!==void 0&&(this.#e=t.kFold),t.nRepeats!==void 0&&(this.#r=t.nRepeats),t.seed!==void 0&&(this.#o=t.seed),this}get capabilities(){return{classifier:this.#n==="classification",regressor:this.#n==="regression",predictProba:this.#n==="classification",decisionFunction:!1,sampleWeight:!1,csr:!1,earlyStopping:!1}}get isFitted(){return this.#h&&!this.#d}get classes(){return this.#i}static _register(){if(Je)return;Je=!0;let t=(n,e,o)=>s._loadFromParts(n,e,o);Ve(Ce,t),Ve(We,t)}static async _loadFromParts(t,n,e){let o=t.params,r=new s({task:o.task,kFold:o.kFold||5,nRepeats:o.nRepeats||1,seed:o.seed??42});r.#i=o.classes?new Int32Array(o.classes):null,r.#f=o.nClasses||0,r.#l=o.nSamples||0,r.#t=[o.estimatorName||"base",null,null];let i=r.#e*r.#r;r.#s=[];for(let f=0;f<i;f++){let h=`fold_${f}`,c=n.find(d=>d.id===h);if(!c)throw new Ye(`No artifact for "${h}"`);let l=e.subarray(c.offset,c.offset+c.length);r.#s.push(await pr(l))}let a=n.find(f=>f.id==="oof");if(a){let f=e.subarray(a.offset,a.offset+a.length),h=new Float64Array(f.buffer.slice(f.byteOffset,f.byteOffset+f.byteLength));r.#a=h,r.#c=new Uint8Array(r.#l).fill(1)}else r.#n==="classification"?r.#a=new Float64Array(r.#l*r.#f):r.#a=new Float64Array(r.#l),r.#c=new Uint8Array(r.#l);return r.#h=!0,r}};function Ge(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function Ar(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}Ke.exports={BaggedEstimator:Pt}});var Nt=x((fo,Qe)=>{var{ValidationError:Ze}=R();function vt(s){let t=s.length;if(t===0)return new Float64Array(0);if(t===1)return new Float64Array([1]);let n=new Float64Array(s);n.sort(),n.reverse();let e=new Float64Array(t);e[0]=n[0];for(let a=1;a<t;a++)e[a]=e[a-1]+n[a];let o=-1;for(let a=0;a<t;a++)n[a]*(a+1)>e[a]-1&&(o=a);if(o<0){let a=new Float64Array(t);return a.fill(1/t),a}let r=(e[o]-1)/(o+1),i=new Float64Array(t);for(let a=0;a<t;a++)i[a]=Math.max(s[a]-r,0);return i}function Fr(s,t,n,{task:e="classification",lr:o=.05,nIter:r=100}={}){let i=s.length,a=t.length;if(i===0)throw new Ze("optimizeWeights: need at least 1 model");if(i===1)return new Float64Array([1]);let f=new Float64Array(n),h=1e-15;if(e==="classification"){let l=s[0].length/a;if(l!==Math.floor(l))throw new Ze("optimizeWeights: prediction length must be divisible by n");for(let d=0;d<r;d++){let u=new Float64Array(i);for(let m=0;m<a;m++){let w=t[m]|0,b=0;for(let g=0;g<i;g++)b+=f[g]*s[g][m*l+w];b=Math.max(b,h);for(let g=0;g<i;g++)u[g]-=s[g][m*l+w]/b}for(let m=0;m<i;m++)u[m]/=a;for(let m=0;m<i;m++)f[m]-=o*u[m];let p=vt(f);for(let m=0;m<i;m++)f[m]=p[m]}}else for(let c=0;c<r;c++){let l=new Float64Array(i);for(let u=0;u<a;u++){let p=0;for(let w=0;w<i;w++)p+=f[w]*s[w][u];let m=t[u]-p;for(let w=0;w<i;w++)l[w]-=2*m*s[w][u]}for(let u=0;u<i;u++)l[u]/=a;for(let u=0;u<i;u++)f[u]-=o*l[u];let d=vt(f);for(let u=0;u<i;u++)f[u]=d[u]}return f}Qe.exports={optimizeWeights:Fr,projectSimplex:vt}});var ss=x((ho,es)=>{var{getScorer:Sr,normalizeY:Er,ValidationError:ts}=R(),{optimizeWeights:Ir}=Nt();function Pr(s,t,{maxSize:n=20,scoring:e="accuracy",task:o="classification",nClasses:r=0,refineWeights:i=!0}={}){let a=Er(t),f=a.length,h=s.length;if(h===0)throw new ts("caruanaSelect: need at least 1 candidate");let c=Sr(e),l=s[0].length/f;if(l!==Math.floor(l))throw new ts("caruanaSelect: oofPredictions[0].length must be divisible by n");o==="classification"&&r===0&&(r=l);let d=new Float64Array(s[0].length),u=[],p=[];for(let y=0;y<n;y++){let F=-1,E=-1/0;for(let S=0;S<h;S++){let A=vr(d,s[S],y,y+1),j=Nr(A,a,c,o,r,f);j>E&&(E=j,F=S)}u.push(F),p.push(E);let v=s[F];for(let S=0;S<d.length;S++)d[S]=(y*d[S]+v[S])/(y+1)}let m=new Map;for(let y of u)m.set(y,(m.get(y)||0)+1);let w=new Int32Array([...m.keys()].sort((y,F)=>y-F)),b=new Float64Array(w.length);for(let y=0;y<w.length;y++)b[y]=m.get(w[y])/n;let g={indices:w,weights:b,scores:new Float64Array(p)};if(i&&w.length>1){let y=Array.from(w,F=>s[F]);g.weights=Ir(y,a,b,{task:o})}return g}function vr(s,t,n,e){let o=new Float64Array(s.length);for(let r=0;r<s.length;r++)o[r]=(n*s[r]+t[r])/e;return o}function Nr(s,t,n,e,o,r){if(e==="regression")return n(t,s);let i=new Float64Array(r);for(let a=0;a<r;a++){let f=0,h=-1/0;for(let c=0;c<o;c++)s[a*o+c]>h&&(h=s[a*o+c],f=c);i[a]=f}return n(t,i)}es.exports={caruanaSelect:Pr}});var os=x((po,rs)=>{var{stratifiedKFold:kr,kFold:Mr,normalizeX:qr,normalizeY:_r,ValidationError:uo}=R();async function $r(s,t,n,{cv:e=5,seed:o=42,task:r="classification"}={}){let i=qr(t),a=_r(n),f=i.rows,h=r==="classification"?kr(a,e,{shuffle:!0,seed:o}):Mr(f,e,{shuffle:!0,seed:o}),c=null,l=0;if(r==="classification"){let u=new Set;for(let p=0;p<a.length;p++)u.add(a[p]);c=new Int32Array([...u].sort((p,m)=>p-m)),l=c.length}let d=[];for(let[u,p,m]of s){let w;r==="classification"?w=new Float64Array(f*l):w=new Float64Array(f);for(let{train:b,test:g}of h){let y=ns(i,b),F=Or(a,b),E=ns(i,g),v=await p.create(m||{});try{if(v.fit(y,F),r==="classification"){let S=await v.predictProba(E);for(let A=0;A<g.length;A++){let j=g[A];for(let _=0;_<l;_++)w[j*l+_]=S[A*l+_]}}else{let S=await v.predict(E);for(let A=0;A<g.length;A++)w[g[A]]=S[A]}}finally{v.dispose()}}d.push(w)}return{oofPreds:d,classes:c}}function ns(s,t){let{data:n,cols:e}=s,o=t.length,r=new Float64Array(o*e);for(let i=0;i<o;i++){let a=t[i]*e;r.set(n.subarray(a,a+e),i*e)}return{data:r,rows:o,cols:e}}function Or(s,t){let n=new s.constructor(t.length);for(let e=0;e<t.length;e++)n[e]=s[t[e]];return n}rs.exports={getOofPredictions:$r}});var Vr=x((mo,is)=>{var{VotingEnsemble:Lr}=qe(),{StackingEnsemble:Br}=Ue(),{BaggedEstimator:Dr}=He(),{caruanaSelect:zr}=ss(),{getOofPredictions:Rr}=os(),{optimizeWeights:jr,projectSimplex:Ur}=Nt();is.exports={VotingEnsemble:Lr,StackingEnsemble:Br,BaggedEstimator:Dr,caruanaSelect:zr,getOofPredictions:Rr,optimizeWeights:jr,projectSimplex:Ur}});return Vr();})();
+var {VotingEnsemble,StackingEnsemble,BaggedEstimator,caruanaSelect,getOofPredictions,optimizeWeights,projectSimplex}=__ensemble;export{VotingEnsemble,StackingEnsemble,BaggedEstimator,caruanaSelect,getOofPredictions,optimizeWeights,projectSimplex};

package/package.json

@@ -1,26 +1,55 @@
 {
   "name": "@wlearn/ensemble",
-  "version": "0.1.0",
+  "version": "0.2.0",
   "description": "Ensemble methods for wlearn: voting, stacking, Caruana selection",
-  "type": "module",
+  "type": "commonjs",
   "main": "src/index.js",
   "exports": {
-    ".": "./src/index.js"
+    ".": {
+      "require": "./src/index.js",
+      "default": "./src/index.js"
+    }
   },
   "files": [
-    "src/"
+    "src/",
+    "dist/",
+    "LICENSE",
+    "NOTICE",
+    "README.md",
+    "CHANGELOG.md"
   ],
   "sideEffects": false,
   "dependencies": {
-    "@wlearn/core": "0.1.0",
-    "@wlearn/types": "0.1.0"
+    "@wlearn/core": "0.2.0"
   },
   "scripts": {
-    "test": "node --test test/*.js"
+    "test": "node --test test/*.js",
+    "build:browser": "bash scripts/build-browser.sh",
+    "prepack": "node scripts/prepack.js"
   },
   "publishConfig": {
     "access": "public"
   },
+  "keywords": [
+    "wlearn",
+    "ensemble",
+    "machine-learning",
+    "stacking",
+    "voting"
+  ],
   "author": "Anton Zemlyansky",
-  "license": "MIT"
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/wlearn-org/wlearn.git",
+    "directory": "packages/ensemble"
+  },
+  "license": "Apache-2.0",
+  "devDependencies": {
+    "esbuild": "^0.27.3",
+    "playwright": "^1.58.2"
+  },
+  "bugs": {
+    "url": "https://github.com/wlearn-org/wlearn/issues"
+  },
+  "homepage": "https://wlearn.org"
 }

package/src/bagging.js

@@ -1,10 +1,10 @@
-import {
-  encodeBundle, decodeBundle, register, load as registryLoad,
+const {
+  encodeBundle, decodeBundle, register, load: registryLoad,
   normalizeX, normalizeY, accuracy, r2Score,
   stratifiedKFold, kFold,
   ValidationError, NotFittedError, DisposedError,
   lift
-} from '@wlearn/core'
+} = require('@wlearn/core')
 
 const TYPE_ID_CLS = 'wlearn.ensemble.bagged.classifier@1'
 const TYPE_ID_REG = 'wlearn.ensemble.bagged.regressor@1'
@@ -17,7 +17,7 @@ let _registered = false
  * seed for fold assignment. OOF predictions are accumulated (sum + count)
  * and averaged, matching AutoGluon's BaggedEnsembleModel pattern.
  */
-export class BaggedEstimator {
+class BaggedEstimator {
   #spec         // [name, Class, params]
   #kFold
   #nRepeats
@@ -408,3 +408,5 @@ function _subsetY(y, indices) {
   }
   return out
 }
+
+module.exports = { BaggedEstimator }

package/src/index.js

@@ -1,6 +1,16 @@
-export { VotingEnsemble } from './voting.js'
-export { StackingEnsemble } from './stacking.js'
-export { BaggedEstimator } from './bagging.js'
-export { caruanaSelect } from './selection.js'
-export { getOofPredictions } from './oof.js'
-export { optimizeWeights, projectSimplex } from './weights.js'
+const { VotingEnsemble } = require('./voting.js')
+const { StackingEnsemble } = require('./stacking.js')
+const { BaggedEstimator } = require('./bagging.js')
+const { caruanaSelect } = require('./selection.js')
+const { getOofPredictions } = require('./oof.js')
+const { optimizeWeights, projectSimplex } = require('./weights.js')
+
+module.exports = {
+  VotingEnsemble,
+  StackingEnsemble,
+  BaggedEstimator,
+  caruanaSelect,
+  getOofPredictions,
+  optimizeWeights,
+  projectSimplex,
+}

package/src/oof.js

@@ -1,15 +1,9 @@
-import { stratifiedKFold, kFold, normalizeX, normalizeY, ValidationError } from '@wlearn/core'
+const { stratifiedKFold, kFold, normalizeX, normalizeY, ValidationError } = require('@wlearn/core')
 
 /**
  * Generate out-of-fold predictions for a list of estimator specs.
- *
- * @param {Array<[string, Function, Object]>} estimatorSpecs - [name, EstimatorClass, params]
- * @param {Object|number[][]} X - feature matrix
- * @param {TypedArray|number[]} y - labels
- * @param {Object} opts
- * @returns {{ oofPreds: Float64Array[], classes: Int32Array|null }}
  */
-export async function getOofPredictions(estimatorSpecs, X, y, {
+async function getOofPredictions(estimatorSpecs, X, y, {
   cv = 5,
   seed = 42,
   task = 'classification',
@@ -94,3 +88,5 @@ function _subsetY(y, indices) {
   }
   return out
 }
+
+module.exports = { getOofPredictions }

package/src/selection.js

@@ -1,22 +1,10 @@
-import { getScorer, normalizeY, ValidationError } from '@wlearn/core'
-import { optimizeWeights } from './weights.js'
+const { getScorer, normalizeY, ValidationError } = require('@wlearn/core')
+const { optimizeWeights } = require('./weights.js')
 
 /**
  * Caruana greedy ensemble selection (Caruana et al., 2004).
- *
- * Selects a weighted subset from a pool of OOF predictions by greedily
- * adding the candidate that most improves the ensemble score at each step.
- * Candidates can be selected multiple times (with replacement).
- *
- * @param {Float64Array[]} oofPredictions - per-candidate OOF predictions
- *   Classification: each is n_samples * n_classes (flat row-major proba)
- *   Regression: each is n_samples
- * @param {TypedArray|number[]} yTrue - true labels
- * @param {Object} opts
- * @param {boolean} opts.refineWeights - if true, optimize weights after selection
- * @returns {{ indices: Int32Array, weights: Float64Array, scores: Float64Array }}
  */
-export function caruanaSelect(oofPredictions, yTrue, {
+function caruanaSelect(oofPredictions, yTrue, {
   maxSize = 20,
   scoring = 'accuracy',
   task = 'classification',
@@ -125,3 +113,5 @@ function _score(preds, yTrue, scorerFn, task, nClasses, n) {
   }
   return scorerFn(yTrue, hardPreds)
 }
+
+module.exports = { caruanaSelect }

package/src/stacking.js

@@ -1,16 +1,16 @@
-import {
-  encodeBundle, decodeBundle, register, load as registryLoad,
+const {
+  encodeBundle, decodeBundle, register, load: registryLoad,
   normalizeX, normalizeY, accuracy, r2Score,
   stratifiedKFold, kFold,
   ValidationError, NotFittedError, DisposedError,
   lift
-} from '@wlearn/core'
+} = require('@wlearn/core')
 
 const TYPE_ID_CLS = 'wlearn.ensemble.stacking.classifier@1'
 const TYPE_ID_REG = 'wlearn.ensemble.stacking.regressor@1'
 let _registered = false
 
-export class StackingEnsemble {
+class StackingEnsemble {
   #baseSpecs      // [name, Class, params][]
   #metaSpec       // [name, Class, params]
   #baseModels     // fitted base model instances (on full data)
@@ -370,3 +370,5 @@ function _subsetY(y, indices) {
   }
   return out
 }
+
+module.exports = { StackingEnsemble }

package/src/voting.js

@@ -1,15 +1,15 @@
-import {
-  encodeBundle, decodeBundle, register, load as registryLoad,
+const {
+  encodeBundle, decodeBundle, register, load: registryLoad,
   normalizeX, normalizeY, accuracy, r2Score,
   ValidationError, NotFittedError, DisposedError,
   lift
-} from '@wlearn/core'
+} = require('@wlearn/core')
 
 const TYPE_ID_CLS = 'wlearn.ensemble.voting.classifier@1'
 const TYPE_ID_REG = 'wlearn.ensemble.voting.regressor@1'
 let _registered = false
 
-export class VotingEnsemble {
+class VotingEnsemble {
   #specs       // [name, Class, params][]
   #models      // fitted instances
   #weights
@@ -309,3 +309,5 @@ export class VotingEnsemble {
     return ens
   }
 }
+
+module.exports = { VotingEnsemble }

package/src/weights.js

@@ -1,13 +1,10 @@
-import { ValidationError } from '@wlearn/core'
+const { ValidationError } = require('@wlearn/core')
 
 /**
  * Project vector onto the probability simplex {w: w >= 0, sum(w) = 1}.
  * O(n log n) algorithm from Duchi et al. 2008.
- *
- * @param {Float64Array} v - input vector
- * @returns {Float64Array} - projected vector
  */
-export function projectSimplex(v) {
+function projectSimplex(v) {
   const n = v.length
   if (n === 0) return new Float64Array(0)
   if (n === 1) return new Float64Array([1.0])
@@ -47,17 +44,8 @@ export function projectSimplex(v) {
 
 /**
  * Optimize ensemble weights via projected gradient descent on the simplex.
- *
- * Classification: minimizes negative log-loss.
- * Regression: minimizes MSE.
- *
- * @param {Float64Array[]} oofPredictions - per-model OOF predictions
- * @param {TypedArray} yTrue - true labels
- * @param {Float64Array} initWeights - initial weights (e.g. from Caruana)
- * @param {object} opts
- * @returns {Float64Array} - optimized weights (>= 0, sum = 1)
  */
-export function optimizeWeights(oofPredictions, yTrue, initWeights, {
+function optimizeWeights(oofPredictions, yTrue, initWeights, {
   task = 'classification',
   lr = 0.05,
   nIter = 100,
@@ -141,3 +129,5 @@ export function optimizeWeights(oofPredictions, yTrue, initWeights, {
 
   return w
 }
+
+module.exports = { optimizeWeights, projectSimplex }