@seanhogg/builderforce-memory-engine 2026.9.34 → 2026.10.1

package/dist/ml/classical/gbm.js

@@ -0,0 +1,116 @@
+import { featureDim, sigmoid } from "./math.js";
+function leafValue(rows, lambda) {
+    let g = 0;
+    let h = 0;
+    for (const r of rows) {
+        g += r.w * r.g;
+        h += r.w * r.h;
+    }
+    return -g / (h + lambda);
+}
+/** XGBoost-style structure score G²/(H+λ) of a node. */
+function gain(g, h, lambda) {
+    return (g * g) / (h + lambda);
+}
+function buildTree(rows, depth, opts) {
+    if (depth >= opts.maxDepth || rows.length < 2)
+        return { value: leafValue(rows, opts.lambda) };
+    let G = 0;
+    let H = 0;
+    let W = 0;
+    for (const r of rows) {
+        G += r.w * r.g;
+        H += r.w * r.h;
+        W += r.w;
+    }
+    const parent = gain(G, H, opts.lambda);
+    let best = null;
+    const d = rows[0].x.length;
+    for (let f = 0; f < d; f++) {
+        const sorted = [...rows].sort((a, b) => a.x[f] - b.x[f]);
+        let gl = 0;
+        let hl = 0;
+        let wl = 0;
+        for (let i = 0; i < sorted.length - 1; i++) {
+            const r = sorted[i];
+            gl += r.w * r.g;
+            hl += r.w * r.h;
+            wl += r.w;
+            const a = r.x[f];
+            const b = sorted[i + 1].x[f];
+            if (a === b)
+                continue;
+            if (wl < opts.minLeafWeight || W - wl < opts.minLeafWeight)
+                continue;
+            const score = gain(gl, hl, opts.lambda) + gain(G - gl, H - hl, opts.lambda) - parent;
+            if (score > 1e-12 && (!best || score > best.score))
+                best = { feature: f, split: (a + b) / 2, score };
+        }
+    }
+    if (!best)
+        return { value: leafValue(rows, opts.lambda) };
+    const left = rows.filter((r) => r.x[best.feature] < best.split);
+    const right = rows.filter((r) => r.x[best.feature] >= best.split);
+    return { feature: best.feature, split: best.split, left: buildTree(left, depth + 1, opts), right: buildTree(right, depth + 1, opts) };
+}
+function treeValue(t, x) {
+    while (!("value" in t))
+        t = x[t.feature] < t.split ? t.left : t.right;
+    return t.value;
+}
+export class GbmModel {
+    objective;
+    base;
+    trees;
+    learningRate;
+    constructor(objective, base, trees, learningRate) {
+        this.objective = objective;
+        this.base = base;
+        this.trees = trees;
+        this.learningRate = learningRate;
+    }
+    /** The raw additive score (log-odds for `binary`). */
+    margin(x) {
+        let s = this.base;
+        for (const t of this.trees)
+            s += this.learningRate * treeValue(t, x);
+        return s;
+    }
+    predict(x) {
+        const m = this.margin(x);
+        if (this.objective === "regression")
+            return { scores: [m], confidence: 1 };
+        const p = sigmoid(m);
+        return { scores: [1 - p, p], confidence: Math.max(p, 1 - p) };
+    }
+}
+export function fitGbm(samples, opts = {}) {
+    featureDim(samples.map((s) => s.x));
+    const objective = opts.objective ?? "binary";
+    const lr = opts.learningRate ?? 0.3;
+    const treeOpts = { maxDepth: opts.maxDepth ?? 2, minLeafWeight: opts.minLeafWeight ?? 1, lambda: opts.lambda ?? 1 };
+    let wSum = 0;
+    let ySum = 0;
+    for (const s of samples) {
+        wSum += s.weight ?? 1;
+        ySum += (s.weight ?? 1) * s.y;
+    }
+    const yMean = ySum / wSum;
+    const base = objective === "binary" ? Math.log(Math.max(yMean, 1e-6) / Math.max(1 - yMean, 1e-6)) : yMean;
+    const margins = samples.map(() => base);
+    const trees = [];
+    for (let t = 0; t < (opts.trees ?? 50); t++) {
+        const rows = samples.map((s, i) => {
+            if (objective === "binary") {
+                const p = sigmoid(margins[i]);
+                return { x: s.x, g: p - s.y, h: Math.max(p * (1 - p), 1e-6), w: s.weight ?? 1 };
+            }
+            return { x: s.x, g: margins[i] - s.y, h: 1, w: s.weight ?? 1 };
+        });
+        const tree = buildTree(rows, 0, treeOpts);
+        trees.push(tree);
+        samples.forEach((s, i) => (margins[i] = margins[i] + lr * treeValue(tree, s.x)));
+    }
+    return new GbmModel(objective, base, trees, lr);
+}
+//# sourceMappingURL=gbm.js.map

package/dist/ml/classical/gbm.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"gbm.js","sourceRoot":"","sources":["../../../src/ml/classical/gbm.ts"],"names":[],"mappings":"AASA,OAAO,EAAE,UAAU,EAAE,OAAO,EAAY,MAAM,WAAW,CAAC;AAwB1D,SAAS,SAAS,CAAC,IAAoB,EAAE,MAAc;IACrD,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,KAAK,MAAM,CAAC,IAAI,IAAI,EAAE,CAAC;QACrB,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QACf,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;IACjB,CAAC;IACD,OAAO,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC;AAC3B,CAAC;AAED,wDAAwD;AACxD,SAAS,IAAI,CAAC,CAAS,EAAE,CAAS,EAAE,MAAc;IAChD,OAAO,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC;AAChC,CAAC;AAED,SAAS,SAAS,CAAC,IAAW,EAAE,KAAa,EAAE,IAAwE;IACrH,IAAI,KAAK,IAAI,IAAI,CAAC,QAAQ,IAAI,IAAI,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,EAAE,KAAK,EAAE,SAAS,CAAC,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;IAC9F,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,IAAI,CAAC,GAAG,CAAC,CAAC;IACV,KAAK,MAAM,CAAC,IAAI,IAAI,EAAE,CAAC;QACrB,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QACf,CAAC,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;QACf,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;IACX,CAAC;IACD,MAAM,MAAM,GAAG,IAAI,CAAC,CAAC,EAAE,CAAC,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;IACvC,IAAI,IAAI,GAA6D,IAAI,CAAC;IAC1E,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC,CAAE,CAAC,CAAC,CAAC,MAAM,CAAC;IAC5B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAC3B,MAAM,MAAM,GAAG,CAAC,GAAG,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAE,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC,CAAC;QAC3D,IAAI,EAAE,GAAG,CAAC,CAAC;QACX,IAAI,EAAE,GAAG,CAAC,CAAC;QACX,IAAI,EAAE,GAAG,CAAC,CAAC;QACX,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;YAC3C,MAAM,CAAC,GAAG,MAAM,CAAC,CAAC,CAAE,CAAC;YACrB,EAAE,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAChB,EAAE,IAAI,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAChB,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC;YACV,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC;YAClB,MAAM,CAAC,GAAG,MAAM,CAAC,CAAC,GAAG,CAAC,CAAE,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC;YAC/B,IAAI,CAAC,KAAK,CAAC;gBAAE,SAAS;YACtB,IAAI,EAAE,GAAG,IAAI,CAAC,aAAa,IAAI,CAAC,GAAG,EAAE,GAAG,IAAI,CAAC,aAAa;gBAAE,SAAS;YACrE,MAAM,KAAK,GAAG,IAAI,CAAC,EAAE,EAAE,EAAE,EAAE,IAAI,CAAC,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,GAAG,EAAE,EAAE,IAAI,CAAC,MAAM,CAAC,GAAG,MAAM,CAAC;YACrF,IAAI,KAAK,GAAG,KAAK,IAAI,CAAC,CAAC,IAAI,IAAI,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC;gBAAE,IAAI,GAAG,EAAE,OAAO,EAAE,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,EAAE,KAAK,EAAE,CAAC;QACvG,CAAC;IACH,CAAC;IACD,IAAI,CAAC,IAAI;QAAE,OAAO,EAAE,KAAK,EAAE,SAAS,CAAC,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,EAAE,CAAC;IAC1D,MAAM,IAAI,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAK,CAAC,OAAO,CAAE,GAAG,IAAK,CAAC,KAAK,CAAC,CAAC;IACnE,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAK,CAAC,OAAO,CAAE,IAAI,IAAK,CAAC,KAAK,CAAC,CAAC;IACrE,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,KAAK,EAAE,IAAI,CAAC,KAAK,EAAE,IAAI,EAAE,SAAS,CAAC,IAAI,EAAE,KAAK,GAAG,CAAC,EAAE,IAAI,CAAC,EAAE,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,KAAK,GAAG,CAAC,EAAE,IAAI,CAAC,EAAE,CAAC;AACxI,CAAC;AAED,SAAS,SAAS,CAAC,CAAQ,EAAE,CAAM;IACjC,OAAO,CAAC,CAAC,OAAO,IAAI,CAAC,CAAC;QAAE,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAE,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;IACvE,OAAO,CAAC,CAAC,KAAK,CAAC;AACjB,CAAC;AAED,MAAM,OAAO,QAAQ;IAER;IACA;IACA;IACA;IAJX,YACW,SAAuB,EACvB,IAAY,EACZ,KAAuB,EACvB,YAAoB;QAHpB,cAAS,GAAT,SAAS,CAAc;QACvB,SAAI,GAAJ,IAAI,CAAQ;QACZ,UAAK,GAAL,KAAK,CAAkB;QACvB,iBAAY,GAAZ,YAAY,CAAQ;IAC5B,CAAC;IAEJ,sDAAsD;IACtD,MAAM,CAAC,CAAM;QACX,IAAI,CAAC,GAAG,IAAI,CAAC,IAAI,CAAC;QAClB,KAAK,MAAM,CAAC,IAAI,IAAI,CAAC,KAAK;YAAE,CAAC,IAAI,IAAI,CAAC,YAAY,GAAG,SAAS,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QACrE,OAAO,CAAC,CAAC;IACX,CAAC;IAED,OAAO,CAAC,CAAe;QACrB,MAAM,CAAC,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;QACzB,IAAI,IAAI,CAAC,SAAS,KAAK,YAAY;YAAE,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC;QAC3E,MAAM,CAAC,GAAG,OAAO,CAAC,CAAC,CAAC,CAAC;QACrB,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC;IAChE,CAAC;CACF;AAED,MAAM,UAAU,MAAM,CAAC,OAA4B,EAAE,OAAmB,EAAE;IACxE,UAAU,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACpC,MAAM,SAAS,GAAG,IAAI,CAAC,SAAS,IAAI,QAAQ,CAAC;IAC7C,MAAM,EAAE,GAAG,IAAI,CAAC,YAAY,IAAI,GAAG,CAAC;IACpC,MAAM,QAAQ,GAAG,EAAE,QAAQ,EAAE,IAAI,CAAC,QAAQ,IAAI,CAAC,EAAE,aAAa,EAAE,IAAI,CAAC,aAAa,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,IAAI,CAAC,EAAE,CAAC;IAEpH,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,IAAI,IAAI,GAAG,CAAC,CAAC;IACb,KAAK,MAAM,CAAC,IAAI,OAAO,EAAE,CAAC;QACxB,IAAI,IAAI,CAAC,CAAC,MAAM,IAAI,CAAC,CAAC;QACtB,IAAI,IAAI,CAAC,CAAC,CAAC,MAAM,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;IAChC,CAAC;IACD,MAAM,KAAK,GAAG,IAAI,GAAG,IAAI,CAAC;IAC1B,MAAM,IAAI,GAAG,SAAS,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,IAAI,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,KAAK,EAAE,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC;IAC1G,MAAM,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,CAAC,CAAC;IACxC,MAAM,KAAK,GAAY,EAAE,CAAC;IAC1B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,IAAI,CAAC,KAAK,IAAI,EAAE,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAC5C,MAAM,IAAI,GAAU,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;YACvC,IAAI,SAAS,KAAK,QAAQ,EAAE,CAAC;gBAC3B,MAAM,CAAC,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC,CAAE,CAAC,CAAC;gBAC/B,OAAO,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,EAAE,CAAC;YAClF,CAAC;YACD,OAAO,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,OAAO,CAAC,CAAC,CAAE,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,EAAE,CAAC;QAClE,CAAC,CAAC,CAAC;QACH,MAAM,IAAI,GAAG,SAAS,CAAC,IAAI,EAAE,CAAC,EAAE,QAAQ,CAAC,CAAC;QAC1C,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACjB,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,OAAO,CAAC,CAAC,CAAE,GAAG,EAAE,GAAG,SAAS,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACpF,CAAC;IACD,OAAO,IAAI,QAAQ,CAAC,SAAS,EAAE,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC;AAClD,CAAC"}

package/dist/ml/classical/index.d.ts

@@ -0,0 +1,48 @@
+/**
+ * ml/classical — pure-TypeScript algorithms, one file each, and the ONE dispatch
+ * from a {@link ClassicalAlgorithm} name to its fit.
+ */
+import type { ClassicalAlgorithm, MlPredictor, MlSample } from "../types.js";
+import { type KnnOptions } from "./knn.js";
+import { type NaiveBayesOptions } from "./naive_bayes.js";
+import { type LogisticOptions } from "./logistic.js";
+import { type LinearOptions } from "./linear.js";
+import { type KMeansOptions } from "./kmeans.js";
+import { type RobustZOptions } from "./robust_z.js";
+import { type EwmaOptions } from "./ewma.js";
+import { type IsolationForestOptions } from "./isolation_forest.js";
+import { type GbmOptions } from "./gbm.js";
+export { KnnModel, fitKnn } from "./knn.js";
+export type { KnnOptions, KnnMetric, Neighbour } from "./knn.js";
+export { NaiveBayesModel, fitNaiveBayes } from "./naive_bayes.js";
+export type { NaiveBayesOptions } from "./naive_bayes.js";
+export { LogisticModel, fitLogistic } from "./logistic.js";
+export type { LogisticOptions } from "./logistic.js";
+export { LinearModel, fitLinear, solveLinearSystem } from "./linear.js";
+export type { LinearOptions } from "./linear.js";
+export { KMeansModel, fitKMeans } from "./kmeans.js";
+export type { KMeansOptions } from "./kmeans.js";
+export { RobustZModel, fitRobustZ, anomalyConfidence } from "./robust_z.js";
+export type { RobustZOptions } from "./robust_z.js";
+export { EwmaModel, fitEwma } from "./ewma.js";
+export type { EwmaOptions } from "./ewma.js";
+export { IsolationForestModel, fitIsolationForest } from "./isolation_forest.js";
+export type { IsolationForestOptions } from "./isolation_forest.js";
+export { GbmModel, fitGbm } from "./gbm.js";
+export type { GbmOptions, GbmObjective } from "./gbm.js";
+export { cosine, dot } from "./math.js";
+/** Per-algorithm options, keyed by algorithm name. */
+export interface ClassicalOptions {
+    knn?: KnnOptions;
+    "naive-bayes"?: NaiveBayesOptions;
+    logistic?: LogisticOptions;
+    linear?: LinearOptions;
+    kmeans?: KMeansOptions;
+    "robust-z"?: RobustZOptions;
+    ewma?: EwmaOptions;
+    "isolation-forest"?: IsolationForestOptions;
+    "gbm-small"?: GbmOptions;
+}
+/** Fit `algorithm` on `samples`. Unsupervised algorithms ignore `y`. */
+export declare function fitClassical(algorithm: ClassicalAlgorithm, samples: readonly MlSample[], opts?: ClassicalOptions): MlPredictor;
+//# sourceMappingURL=index.d.ts.map

package/dist/ml/classical/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/ml/classical/index.ts"],"names":[],"mappings":"AAAA;;;GAGG;AACH,OAAO,KAAK,EAAE,kBAAkB,EAAE,WAAW,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAC7E,OAAO,EAAU,KAAK,UAAU,EAAE,MAAM,UAAU,CAAC;AACnD,OAAO,EAAiB,KAAK,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AACzE,OAAO,EAAe,KAAK,eAAe,EAAE,MAAM,eAAe,CAAC;AAClE,OAAO,EAAa,KAAK,aAAa,EAAE,MAAM,aAAa,CAAC;AAC5D,OAAO,EAAa,KAAK,aAAa,EAAE,MAAM,aAAa,CAAC;AAC5D,OAAO,EAAc,KAAK,cAAc,EAAE,MAAM,eAAe,CAAC;AAChE,OAAO,EAAW,KAAK,WAAW,EAAE,MAAM,WAAW,CAAC;AACtD,OAAO,EAAsB,KAAK,sBAAsB,EAAE,MAAM,uBAAuB,CAAC;AACxF,OAAO,EAAU,KAAK,UAAU,EAAE,MAAM,UAAU,CAAC;AAEnD,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM,UAAU,CAAC;AAC5C,YAAY,EAAE,UAAU,EAAE,SAAS,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AACjE,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAClE,YAAY,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAC1D,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC3D,YAAY,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AACrD,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AACxE,YAAY,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AACjD,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AACrD,YAAY,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AACjD,OAAO,EAAE,YAAY,EAAE,UAAU,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AAC5E,YAAY,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,EAAE,SAAS,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAC/C,YAAY,EAAE,WAAW,EAAE,MAAM,WAAW,CAAC;AAC7C,OAAO,EAAE,oBAAoB,EAAE,kBAAkB,EAAE,MAAM,uBAAuB,CAAC;AACjF,YAAY,EAAE,sBAAsB,EAAE,MAAM,uBAAuB,CAAC;AACpE,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM,UAAU,CAAC;AAC5C,YAAY,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AACzD,OAAO,EAAE,MAAM,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AAExC,sDAAsD;AACtD,MAAM,WAAW,gBAAgB;IAC/B,GAAG,CAAC,EAAE,UAAU,CAAC;IACjB,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,QAAQ,CAAC,EAAE,eAAe,CAAC;IAC3B,MAAM,CAAC,EAAE,aAAa,CAAC;IACvB,MAAM,CAAC,EAAE,aAAa,CAAC;IACvB,UAAU,CAAC,EAAE,cAAc,CAAC;IAC5B,IAAI,CAAC,EAAE,WAAW,CAAC;IACnB,kBAAkB,CAAC,EAAE,sBAAsB,CAAC;IAC5C,WAAW,CAAC,EAAE,UAAU,CAAC;CAC1B;AAED,wEAAwE;AACxE,wBAAgB,YAAY,CAC1B,SAAS,EAAE,kBAAkB,EAC7B,OAAO,EAAE,SAAS,QAAQ,EAAE,EAC5B,IAAI,GAAE,gBAAqB,GAC1B,WAAW,CAsBb"}

package/dist/ml/classical/index.js

@@ -0,0 +1,45 @@
+import { fitKnn } from "./knn.js";
+import { fitNaiveBayes } from "./naive_bayes.js";
+import { fitLogistic } from "./logistic.js";
+import { fitLinear } from "./linear.js";
+import { fitKMeans } from "./kmeans.js";
+import { fitRobustZ } from "./robust_z.js";
+import { fitEwma } from "./ewma.js";
+import { fitIsolationForest } from "./isolation_forest.js";
+import { fitGbm } from "./gbm.js";
+export { KnnModel, fitKnn } from "./knn.js";
+export { NaiveBayesModel, fitNaiveBayes } from "./naive_bayes.js";
+export { LogisticModel, fitLogistic } from "./logistic.js";
+export { LinearModel, fitLinear, solveLinearSystem } from "./linear.js";
+export { KMeansModel, fitKMeans } from "./kmeans.js";
+export { RobustZModel, fitRobustZ, anomalyConfidence } from "./robust_z.js";
+export { EwmaModel, fitEwma } from "./ewma.js";
+export { IsolationForestModel, fitIsolationForest } from "./isolation_forest.js";
+export { GbmModel, fitGbm } from "./gbm.js";
+export { cosine, dot } from "./math.js";
+/** Fit `algorithm` on `samples`. Unsupervised algorithms ignore `y`. */
+export function fitClassical(algorithm, samples, opts = {}) {
+    if (samples.length === 0)
+        throw new Error(`fitClassical(${algorithm}): no samples`);
+    switch (algorithm) {
+        case "knn":
+            return fitKnn(samples, opts.knn);
+        case "naive-bayes":
+            return fitNaiveBayes(samples, opts["naive-bayes"]);
+        case "logistic":
+            return fitLogistic(samples, opts.logistic);
+        case "linear":
+            return fitLinear(samples, opts.linear);
+        case "kmeans":
+            return fitKMeans(samples, opts.kmeans);
+        case "robust-z":
+            return fitRobustZ(samples, opts["robust-z"]);
+        case "ewma":
+            return fitEwma(samples, opts.ewma);
+        case "isolation-forest":
+            return fitIsolationForest(samples, opts["isolation-forest"]);
+        case "gbm-small":
+            return fitGbm(samples, opts["gbm-small"]);
+    }
+}
+//# sourceMappingURL=index.js.map

package/dist/ml/classical/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/ml/classical/index.ts"],"names":[],"mappings":"AAKA,OAAO,EAAE,MAAM,EAAmB,MAAM,UAAU,CAAC;AACnD,OAAO,EAAE,aAAa,EAA0B,MAAM,kBAAkB,CAAC;AACzE,OAAO,EAAE,WAAW,EAAwB,MAAM,eAAe,CAAC;AAClE,OAAO,EAAE,SAAS,EAAsB,MAAM,aAAa,CAAC;AAC5D,OAAO,EAAE,SAAS,EAAsB,MAAM,aAAa,CAAC;AAC5D,OAAO,EAAE,UAAU,EAAuB,MAAM,eAAe,CAAC;AAChE,OAAO,EAAE,OAAO,EAAoB,MAAM,WAAW,CAAC;AACtD,OAAO,EAAE,kBAAkB,EAA+B,MAAM,uBAAuB,CAAC;AACxF,OAAO,EAAE,MAAM,EAAmB,MAAM,UAAU,CAAC;AAEnD,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM,UAAU,CAAC;AAE5C,OAAO,EAAE,eAAe,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAElE,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAE3D,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,iBAAiB,EAAE,MAAM,aAAa,CAAC;AAExE,OAAO,EAAE,WAAW,EAAE,SAAS,EAAE,MAAM,aAAa,CAAC;AAErD,OAAO,EAAE,YAAY,EAAE,UAAU,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AAE5E,OAAO,EAAE,SAAS,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AAE/C,OAAO,EAAE,oBAAoB,EAAE,kBAAkB,EAAE,MAAM,uBAAuB,CAAC;AAEjF,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,MAAM,UAAU,CAAC;AAE5C,OAAO,EAAE,MAAM,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AAexC,wEAAwE;AACxE,MAAM,UAAU,YAAY,CAC1B,SAA6B,EAC7B,OAA4B,EAC5B,OAAyB,EAAE;IAE3B,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC;QAAE,MAAM,IAAI,KAAK,CAAC,gBAAgB,SAAS,eAAe,CAAC,CAAC;IACpF,QAAQ,SAAS,EAAE,CAAC;QAClB,KAAK,KAAK;YACR,OAAO,MAAM,CAAC,OAAO,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC;QACnC,KAAK,aAAa;YAChB,OAAO,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,aAAa,CAAC,CAAC,CAAC;QACrD,KAAK,UAAU;YACb,OAAO,WAAW,CAAC,OAAO,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC;QAC7C,KAAK,QAAQ;YACX,OAAO,SAAS,CAAC,OAAO,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;QACzC,KAAK,QAAQ;YACX,OAAO,SAAS,CAAC,OAAO,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;QACzC,KAAK,UAAU;YACb,OAAO,UAAU,CAAC,OAAO,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC;QAC/C,KAAK,MAAM;YACT,OAAO,OAAO,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QACrC,KAAK,kBAAkB;YACrB,OAAO,kBAAkB,CAAC,OAAO,EAAE,IAAI,CAAC,kBAAkB,CAAC,CAAC,CAAC;QAC/D,KAAK,WAAW;YACd,OAAO,MAAM,CAAC,OAAO,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC;IAC9C,CAAC;AACH,CAAC"}

package/dist/ml/classical/isolation_forest.d.ts

@@ -0,0 +1,37 @@
+/**
+ * ml/classical/isolation_forest.ts — Isolation Forest (Liu, Ting & Zhou 2008).
+ *
+ * Anomalies are few and different, so random axis-aligned splits isolate them in
+ * fewer steps. Score s(x) = 2^(−E[h(x)] / c(ψ)): near 1 = anomalous, ≤ 0.5 =
+ * normal. Prediction: `scores[0] = s / threshold` — ≥ 1 is anomalous.
+ */
+import type { MlPrediction, MlPredictor, MlSample } from "../types.js";
+import { type Vec } from "./math.js";
+export interface IsolationForestOptions {
+    trees?: number;
+    /** Subsample size ψ per tree. Default min(256, n). */
+    sampleSize?: number;
+    /** Raw score at which a point is anomalous. Default 0.6. */
+    threshold?: number;
+    seed?: number;
+}
+type ITree = {
+    size: number;
+} | {
+    feature: number;
+    split: number;
+    left: ITree;
+    right: ITree;
+};
+export declare class IsolationForestModel implements MlPredictor {
+    readonly trees: readonly ITree[];
+    readonly sampleSize: number;
+    readonly threshold: number;
+    constructor(trees: readonly ITree[], sampleSize: number, threshold: number);
+    /** The raw isolation score in (0, 1]. */
+    score(x: Vec): number;
+    predict(x: Float32Array): MlPrediction;
+}
+export declare function fitIsolationForest(samples: readonly MlSample[], opts?: IsolationForestOptions): IsolationForestModel;
+export {};
+//# sourceMappingURL=isolation_forest.d.ts.map

package/dist/ml/classical/isolation_forest.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"isolation_forest.d.ts","sourceRoot":"","sources":["../../../src/ml/classical/isolation_forest.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,KAAK,EAAE,YAAY,EAAE,WAAW,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvE,OAAO,EAAgD,KAAK,GAAG,EAAE,MAAM,WAAW,CAAC;AAGnF,MAAM,WAAW,sBAAsB;IACrC,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,sDAAsD;IACtD,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,4DAA4D;IAC5D,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,IAAI,CAAC,EAAE,MAAM,CAAC;CACf;AAED,KAAK,KAAK,GAAG;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAAG;IAAE,OAAO,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,KAAK,CAAC;IAAC,KAAK,EAAE,KAAK,CAAA;CAAE,CAAC;AA6B9F,qBAAa,oBAAqB,YAAW,WAAW;IAEpD,QAAQ,CAAC,KAAK,EAAE,SAAS,KAAK,EAAE;IAChC,QAAQ,CAAC,UAAU,EAAE,MAAM;IAC3B,QAAQ,CAAC,SAAS,EAAE,MAAM;gBAFjB,KAAK,EAAE,SAAS,KAAK,EAAE,EACvB,UAAU,EAAE,MAAM,EAClB,SAAS,EAAE,MAAM;IAG5B,yCAAyC;IACzC,KAAK,CAAC,CAAC,EAAE,GAAG,GAAG,MAAM;IAOrB,OAAO,CAAC,CAAC,EAAE,YAAY,GAAG,YAAY;CAIvC;AAED,wBAAgB,kBAAkB,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,EAAE,IAAI,GAAE,sBAA2B,GAAG,oBAAoB,CAiBxH"}

package/dist/ml/classical/isolation_forest.js

@@ -0,0 +1,72 @@
+import { averagePathLength, featureDim, seededUniform } from "./math.js";
+import { anomalyConfidence } from "./robust_z.js";
+function buildTree(X, depth, limit, rnd) {
+    if (depth >= limit || X.length <= 1)
+        return { size: X.length };
+    const d = X[0].length;
+    // Only features that still vary can split; none left = an isolated leaf.
+    const varying = [];
+    for (let f = 0; f < d; f++) {
+        let lo = Infinity;
+        let hi = -Infinity;
+        for (const x of X) {
+            lo = Math.min(lo, x[f]);
+            hi = Math.max(hi, x[f]);
+        }
+        if (hi > lo)
+            varying.push({ f, lo, hi });
+    }
+    if (varying.length === 0)
+        return { size: X.length };
+    const { f, lo, hi } = varying[Math.floor(rnd() * varying.length)];
+    const split = lo + rnd() * (hi - lo);
+    const left = X.filter((x) => x[f] < split);
+    const right = X.filter((x) => x[f] >= split);
+    return { feature: f, split, left: buildTree(left, depth + 1, limit, rnd), right: buildTree(right, depth + 1, limit, rnd) };
+}
+function pathLength(tree, x, depth = 0) {
+    if ("size" in tree)
+        return depth + averagePathLength(tree.size);
+    return pathLength(x[tree.feature] < tree.split ? tree.left : tree.right, x, depth + 1);
+}
+export class IsolationForestModel {
+    trees;
+    sampleSize;
+    threshold;
+    constructor(trees, sampleSize, threshold) {
+        this.trees = trees;
+        this.sampleSize = sampleSize;
+        this.threshold = threshold;
+    }
+    /** The raw isolation score in (0, 1]. */
+    score(x) {
+        const c = averagePathLength(this.sampleSize);
+        if (c === 0)
+            return 0.5;
+        const mean = this.trees.reduce((s, t) => s + pathLength(t, x), 0) / this.trees.length;
+        return Math.pow(2, -mean / c);
+    }
+    predict(x) {
+        const s = this.score(x) / this.threshold;
+        return { scores: [s], confidence: anomalyConfidence(s) };
+    }
+}
+export function fitIsolationForest(samples, opts = {}) {
+    const X = samples.map((s) => s.x);
+    featureDim(X);
+    const rnd = seededUniform(opts.seed ?? 0x15f0);
+    const psi = Math.min(opts.sampleSize ?? 256, X.length);
+    const limit = Math.ceil(Math.log2(Math.max(psi, 2)));
+    const trees = [];
+    for (let t = 0; t < (opts.trees ?? 100); t++) {
+        // Sample ψ points without replacement (partial Fisher–Yates over indices).
+        const idx = X.map((_, i) => i);
+        for (let i = 0; i < psi; i++) {
+            const j = i + Math.floor(rnd() * (idx.length - i));
+            [idx[i], idx[j]] = [idx[j], idx[i]];
+        }
+        trees.push(buildTree(idx.slice(0, psi).map((i) => X[i]), 0, limit, rnd));
+    }
+    return new IsolationForestModel(trees, psi, opts.threshold ?? 0.6);
+}
+//# sourceMappingURL=isolation_forest.js.map

package/dist/ml/classical/isolation_forest.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"isolation_forest.js","sourceRoot":"","sources":["../../../src/ml/classical/isolation_forest.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,iBAAiB,EAAE,UAAU,EAAE,aAAa,EAAY,MAAM,WAAW,CAAC;AACnF,OAAO,EAAE,iBAAiB,EAAE,MAAM,eAAe,CAAC;AAalD,SAAS,SAAS,CAAC,CAAQ,EAAE,KAAa,EAAE,KAAa,EAAE,GAAiB;IAC1E,IAAI,KAAK,IAAI,KAAK,IAAI,CAAC,CAAC,MAAM,IAAI,CAAC;QAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC;IAC/D,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAE,CAAC,MAAM,CAAC;IACvB,yEAAyE;IACzE,MAAM,OAAO,GAA4C,EAAE,CAAC;IAC5D,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAC3B,IAAI,EAAE,GAAG,QAAQ,CAAC;QAClB,IAAI,EAAE,GAAG,CAAC,QAAQ,CAAC;QACnB,KAAK,MAAM,CAAC,IAAI,CAAC,EAAE,CAAC;YAClB,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAE,CAAC,CAAC;YACzB,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAE,CAAC,CAAC;QAC3B,CAAC;QACD,IAAI,EAAE,GAAG,EAAE;YAAE,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC;IAC3C,CAAC;IACD,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC;IACpD,MAAM,EAAE,CAAC,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,OAAO,CAAC,MAAM,CAAC,CAAE,CAAC;IACnE,MAAM,KAAK,GAAG,EAAE,GAAG,GAAG,EAAE,GAAG,CAAC,EAAE,GAAG,EAAE,CAAC,CAAC;IACrC,MAAM,IAAI,GAAG,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAE,GAAG,KAAK,CAAC,CAAC;IAC5C,MAAM,KAAK,GAAG,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAE,IAAI,KAAK,CAAC,CAAC;IAC9C,OAAO,EAAE,OAAO,EAAE,CAAC,EAAE,KAAK,EAAE,IAAI,EAAE,SAAS,CAAC,IAAI,EAAE,KAAK,GAAG,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,EAAE,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,KAAK,GAAG,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,EAAE,CAAC;AAC7H,CAAC;AAED,SAAS,UAAU,CAAC,IAAW,EAAE,CAAM,EAAE,KAAK,GAAG,CAAC;IAChD,IAAI,MAAM,IAAI,IAAI;QAAE,OAAO,KAAK,GAAG,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAChE,OAAO,UAAU,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAE,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,EAAE,CAAC,EAAE,KAAK,GAAG,CAAC,CAAC,CAAC;AAC1F,CAAC;AAED,MAAM,OAAO,oBAAoB;IAEpB;IACA;IACA;IAHX,YACW,KAAuB,EACvB,UAAkB,EAClB,SAAiB;QAFjB,UAAK,GAAL,KAAK,CAAkB;QACvB,eAAU,GAAV,UAAU,CAAQ;QAClB,cAAS,GAAT,SAAS,CAAQ;IACzB,CAAC;IAEJ,yCAAyC;IACzC,KAAK,CAAC,CAAM;QACV,MAAM,CAAC,GAAG,iBAAiB,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;QAC7C,IAAI,CAAC,KAAK,CAAC;YAAE,OAAO,GAAG,CAAC;QACxB,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,UAAU,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC;QACtF,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC;IAChC,CAAC;IAED,OAAO,CAAC,CAAe;QACrB,MAAM,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,SAAS,CAAC;QACzC,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,iBAAiB,CAAC,CAAC,CAAC,EAAE,CAAC;IAC3D,CAAC;CACF;AAED,MAAM,UAAU,kBAAkB,CAAC,OAA4B,EAAE,OAA+B,EAAE;IAChG,MAAM,CAAC,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAClC,UAAU,CAAC,CAAC,CAAC,CAAC;IACd,MAAM,GAAG,GAAG,aAAa,CAAC,IAAI,CAAC,IAAI,IAAI,MAAM,CAAC,CAAC;IAC/C,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,UAAU,IAAI,GAAG,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC;IACvD,MAAM,KAAK,GAAG,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IACrD,MAAM,KAAK,GAAY,EAAE,CAAC;IAC1B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,IAAI,CAAC,KAAK,IAAI,GAAG,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;QAC7C,2EAA2E;QAC3E,MAAM,GAAG,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC;QAC/B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,GAAG,EAAE,CAAC,EAAE,EAAE,CAAC;YAC7B,MAAM,CAAC,GAAG,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC;YACnD,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC,CAAE,EAAE,GAAG,CAAC,CAAC,CAAE,CAAC,CAAC;QACxC,CAAC;QACD,KAAK,CAAC,IAAI,CAAC,SAAS,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAE,CAAC,EAAE,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC;IAC5E,CAAC;IACD,OAAO,IAAI,oBAAoB,CAAC,KAAK,EAAE,GAAG,EAAE,IAAI,CAAC,SAAS,IAAI,GAAG,CAAC,CAAC;AACrE,CAAC"}

package/dist/ml/classical/kmeans.d.ts

@@ -0,0 +1,31 @@
+/**
+ * ml/classical/kmeans.ts — k-means clustering (k-means++ seeding, Lloyd iterations).
+ *
+ * Deterministic for a given seed. A prediction is the nearest centroid; its
+ * confidence is how much closer that centroid is than the runner-up.
+ */
+import type { MlPrediction, MlPredictor, MlSample } from "../types.js";
+import { type Vec } from "./math.js";
+export interface KMeansOptions {
+    /** Clusters. Default round(√(n/2)), at least 1. */
+    k?: number;
+    maxIterations?: number;
+    seed?: number;
+}
+export declare class KMeansModel implements MlPredictor {
+    readonly centroids: Float32Array[];
+    /** Sum of squared distances of the fitted points to their centroid. */
+    readonly inertia: number;
+    constructor(centroids: Float32Array[], 
+    /** Sum of squared distances of the fitted points to their centroid. */
+    inertia: number);
+    /** Nearest centroid index and the two smallest squared distances. */
+    assign(x: Vec): {
+        cluster: number;
+        d1: number;
+        d2: number;
+    };
+    predict(x: Float32Array): MlPrediction;
+}
+export declare function fitKMeans(samples: readonly MlSample[], opts?: KMeansOptions): KMeansModel;
+//# sourceMappingURL=kmeans.d.ts.map

package/dist/ml/classical/kmeans.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"kmeans.d.ts","sourceRoot":"","sources":["../../../src/ml/classical/kmeans.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,YAAY,EAAE,WAAW,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvE,OAAO,EAA8C,KAAK,GAAG,EAAE,MAAM,WAAW,CAAC;AAEjF,MAAM,WAAW,aAAa;IAC5B,mDAAmD;IACnD,CAAC,CAAC,EAAE,MAAM,CAAC;IACX,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,IAAI,CAAC,EAAE,MAAM,CAAC;CACf;AAED,qBAAa,WAAY,YAAW,WAAW;IAE3C,QAAQ,CAAC,SAAS,EAAE,YAAY,EAAE;IAClC,uEAAuE;IACvE,QAAQ,CAAC,OAAO,EAAE,MAAM;gBAFf,SAAS,EAAE,YAAY,EAAE;IAClC,uEAAuE;IAC9D,OAAO,EAAE,MAAM;IAG1B,qEAAqE;IACrE,MAAM,CAAC,CAAC,EAAE,GAAG,GAAG;QAAE,OAAO,EAAE,MAAM,CAAC;QAAC,EAAE,EAAE,MAAM,CAAC;QAAC,EAAE,EAAE,MAAM,CAAA;KAAE;IAe3D,OAAO,CAAC,CAAC,EAAE,YAAY,GAAG,YAAY;CAKvC;AAED,wBAAgB,SAAS,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,EAAE,IAAI,GAAE,aAAkB,GAAG,WAAW,CAuD7F"}

package/dist/ml/classical/kmeans.js

@@ -0,0 +1,93 @@
+import { featureDim, seededUniform, squaredDistance } from "./math.js";
+export class KMeansModel {
+    centroids;
+    inertia;
+    constructor(centroids, 
+    /** Sum of squared distances of the fitted points to their centroid. */
+    inertia) {
+        this.centroids = centroids;
+        this.inertia = inertia;
+    }
+    /** Nearest centroid index and the two smallest squared distances. */
+    assign(x) {
+        let cluster = 0;
+        let d1 = Infinity;
+        let d2 = Infinity;
+        this.centroids.forEach((c, i) => {
+            const d = squaredDistance(x, c);
+            if (d < d1) {
+                d2 = d1;
+                d1 = d;
+                cluster = i;
+            }
+            else if (d < d2)
+                d2 = d;
+        });
+        return { cluster, d1, d2 };
+    }
+    predict(x) {
+        const { cluster, d1, d2 } = this.assign(x);
+        const confidence = Number.isFinite(d2) && d2 > 0 ? 1 - Math.sqrt(d1) / Math.sqrt(d2) : 1;
+        return { scores: [cluster], confidence: Math.max(0, Math.min(1, confidence)) };
+    }
+}
+export function fitKMeans(samples, opts = {}) {
+    const X = samples.map((s) => s.x);
+    const d = featureDim(X);
+    const k = Math.min(X.length, Math.max(1, opts.k ?? Math.round(Math.sqrt(X.length / 2))));
+    const rnd = seededUniform(opts.seed ?? 0x6b6d);
+    // k-means++: first centroid uniformly, the rest ∝ squared distance to the nearest chosen one.
+    const centroids = [Float32Array.from(X[Math.floor(rnd() * X.length)])];
+    const nearest = X.map((x) => squaredDistance(x, centroids[0]));
+    while (centroids.length < k) {
+        const total = nearest.reduce((a, b) => a + b, 0);
+        let pick = 0;
+        if (total > 0) {
+            let r = rnd() * total;
+            for (pick = 0; pick < X.length - 1; pick++) {
+                r -= nearest[pick];
+                if (r <= 0)
+                    break;
+            }
+        }
+        else {
+            pick = centroids.length; // every point coincides with a centroid; take the next index
+        }
+        const c = Float32Array.from(X[pick]);
+        centroids.push(c);
+        X.forEach((x, i) => (nearest[i] = Math.min(nearest[i], squaredDistance(x, c))));
+    }
+    const assign = new Int32Array(X.length).fill(-1);
+    const maxIt = opts.maxIterations ?? 100;
+    for (let it = 0; it < maxIt; it++) {
+        let changed = false;
+        const model = new KMeansModel(centroids, 0);
+        X.forEach((x, i) => {
+            const c = model.assign(x).cluster;
+            if (assign[i] !== c) {
+                assign[i] = c;
+                changed = true;
+            }
+        });
+        if (!changed)
+            break;
+        const sums = centroids.map(() => new Float64Array(d));
+        const counts = new Array(k).fill(0);
+        X.forEach((x, i) => {
+            const c = assign[i];
+            counts[c] = counts[c] + 1;
+            for (let j = 0; j < d; j++)
+                sums[c][j] = sums[c][j] + x[j];
+        });
+        // An emptied cluster keeps its previous centroid.
+        centroids.forEach((c, ci) => {
+            if (counts[ci] > 0)
+                for (let j = 0; j < d; j++)
+                    c[j] = sums[ci][j] / counts[ci];
+        });
+    }
+    let inertia = 0;
+    X.forEach((x, i) => (inertia += squaredDistance(x, centroids[assign[i]])));
+    return new KMeansModel(centroids, inertia);
+}
+//# sourceMappingURL=kmeans.js.map

package/dist/ml/classical/kmeans.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"kmeans.js","sourceRoot":"","sources":["../../../src/ml/classical/kmeans.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,UAAU,EAAE,aAAa,EAAE,eAAe,EAAY,MAAM,WAAW,CAAC;AASjF,MAAM,OAAO,WAAW;IAEX;IAEA;IAHX,YACW,SAAyB;IAClC,uEAAuE;IAC9D,OAAe;QAFf,cAAS,GAAT,SAAS,CAAgB;QAEzB,YAAO,GAAP,OAAO,CAAQ;IACvB,CAAC;IAEJ,qEAAqE;IACrE,MAAM,CAAC,CAAM;QACX,IAAI,OAAO,GAAG,CAAC,CAAC;QAChB,IAAI,EAAE,GAAG,QAAQ,CAAC;QAClB,IAAI,EAAE,GAAG,QAAQ,CAAC;QAClB,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;YAC9B,MAAM,CAAC,GAAG,eAAe,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;YAChC,IAAI,CAAC,GAAG,EAAE,EAAE,CAAC;gBACX,EAAE,GAAG,EAAE,CAAC;gBACR,EAAE,GAAG,CAAC,CAAC;gBACP,OAAO,GAAG,CAAC,CAAC;YACd,CAAC;iBAAM,IAAI,CAAC,GAAG,EAAE;gBAAE,EAAE,GAAG,CAAC,CAAC;QAC5B,CAAC,CAAC,CAAC;QACH,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,EAAE,EAAE,CAAC;IAC7B,CAAC;IAED,OAAO,CAAC,CAAe;QACrB,MAAM,EAAE,OAAO,EAAE,EAAE,EAAE,EAAE,EAAE,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC;QAC3C,MAAM,UAAU,GAAG,MAAM,CAAC,QAAQ,CAAC,EAAE,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QACzF,OAAO,EAAE,MAAM,EAAE,CAAC,OAAO,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,UAAU,CAAC,CAAC,EAAE,CAAC;IACjF,CAAC;CACF;AAED,MAAM,UAAU,SAAS,CAAC,OAA4B,EAAE,OAAsB,EAAE;IAC9E,MAAM,CAAC,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IAClC,MAAM,CAAC,GAAG,UAAU,CAAC,CAAC,CAAC,CAAC;IACxB,MAAM,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,CAAC,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACzF,MAAM,GAAG,GAAG,aAAa,CAAC,IAAI,CAAC,IAAI,IAAI,MAAM,CAAC,CAAC;IAE/C,8FAA8F;IAC9F,MAAM,SAAS,GAAmB,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,MAAM,CAAC,CAAE,CAAC,CAAC,CAAC;IACxF,MAAM,OAAO,GAAG,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,eAAe,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,CAAE,CAAC,CAAC,CAAC;IAChE,OAAO,SAAS,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC5B,MAAM,KAAK,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;QACjD,IAAI,IAAI,GAAG,CAAC,CAAC;QACb,IAAI,KAAK,GAAG,CAAC,EAAE,CAAC;YACd,IAAI,CAAC,GAAG,GAAG,EAAE,GAAG,KAAK,CAAC;YACtB,KAAK,IAAI,GAAG,CAAC,EAAE,IAAI,GAAG,CAAC,CAAC,MAAM,GAAG,CAAC,EAAE,IAAI,EAAE,EAAE,CAAC;gBAC3C,CAAC,IAAI,OAAO,CAAC,IAAI,CAAE,CAAC;gBACpB,IAAI,CAAC,IAAI,CAAC;oBAAE,MAAM;YACpB,CAAC;QACH,CAAC;aAAM,CAAC;YACN,IAAI,GAAG,SAAS,CAAC,MAAM,CAAC,CAAC,6DAA6D;QACxF,CAAC;QACD,MAAM,CAAC,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAE,CAAC,CAAC;QACtC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QAClB,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAE,EAAE,eAAe,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACnF,CAAC;IAED,MAAM,MAAM,GAAG,IAAI,UAAU,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC;IACjD,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,IAAI,GAAG,CAAC;IACxC,KAAK,IAAI,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,KAAK,EAAE,EAAE,EAAE,EAAE,CAAC;QAClC,IAAI,OAAO,GAAG,KAAK,CAAC;QACpB,MAAM,KAAK,GAAG,IAAI,WAAW,CAAC,SAAS,EAAE,CAAC,CAAC,CAAC;QAC5C,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;YACjB,MAAM,CAAC,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC;YAClC,IAAI,MAAM,CAAC,CAAC,CAAC,KAAK,CAAC,EAAE,CAAC;gBACpB,MAAM,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC;gBACd,OAAO,GAAG,IAAI,CAAC;YACjB,CAAC;QACH,CAAC,CAAC,CAAC;QACH,IAAI,CAAC,OAAO;YAAE,MAAM;QACpB,MAAM,IAAI,GAAG,SAAS,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,IAAI,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC;QACtD,MAAM,MAAM,GAAG,IAAI,KAAK,CAAS,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QAC5C,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE;YACjB,MAAM,CAAC,GAAG,MAAM,CAAC,CAAC,CAAE,CAAC;YACrB,MAAM,CAAC,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC,CAAE,GAAG,CAAC,CAAC;YAC3B,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE;gBAAE,IAAI,CAAC,CAAC,CAAE,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC,CAAE,CAAC,CAAC,CAAE,GAAG,CAAC,CAAC,CAAC,CAAE,CAAC;QACjE,CAAC,CAAC,CAAC;QACH,kDAAkD;QAClD,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE;YAC1B,IAAI,MAAM,CAAC,EAAE,CAAE,GAAG,CAAC;gBAAE,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,EAAE;oBAAE,CAAC,CAAC,CAAC,CAAC,GAAG,IAAI,CAAC,EAAE,CAAE,CAAC,CAAC,CAAE,GAAG,MAAM,CAAC,EAAE,CAAE,CAAC;QACtF,CAAC,CAAC,CAAC;IACL,CAAC;IAED,IAAI,OAAO,GAAG,CAAC,CAAC;IAChB,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,OAAO,IAAI,eAAe,CAAC,CAAC,EAAE,SAAS,CAAC,MAAM,CAAC,CAAC,CAAE,CAAE,CAAC,CAAC,CAAC,CAAC;IAC7E,OAAO,IAAI,WAAW,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC;AAC7C,CAAC"}

package/dist/ml/classical/knn.d.ts

@@ -0,0 +1,48 @@
+/**
+ * ml/classical/knn.ts — k-nearest-neighbours classification over feature vectors.
+ *
+ * Votes are similarity-weighted (cosine similarity, or 1/(1+distance) for the
+ * euclidean metric), so a close neighbour counts for more than a far one. The
+ * same index answers "what is my nearest neighbour?" — the de-duplication query.
+ */
+import type { MlPrediction, MlPredictor, MlSample } from "../types.js";
+import { type Vec } from "./math.js";
+export type KnnMetric = "cosine" | "euclidean";
+export interface KnnOptions {
+    /** Neighbours that vote. Default 5 (capped at the sample count). */
+    k?: number;
+    metric?: KnnMetric;
+    /** Number of classes (default: inferred from the labels, at least 2). */
+    classes?: number;
+}
+export interface Neighbour {
+    index: number;
+    /** Higher = closer: cosine similarity, or 1/(1+distance) for euclidean. */
+    similarity: number;
+}
+export declare class KnnModel implements MlPredictor {
+    readonly k: number;
+    readonly metric: KnnMetric;
+    readonly classes: number;
+    private readonly _x;
+    private readonly _y;
+    private readonly _w;
+    constructor(k: number, metric: KnnMetric, classes: number);
+    /** How many points the index holds. */
+    get size(): number;
+    /** Add one labelled point (online — the index never needs a refit). */
+    add(x: Vec, y?: number, weight?: number): void;
+    similarity(a: Vec, b: Vec): number;
+    /** The `k` closest points, closest first. */
+    neighbours(x: Vec, k?: number): Neighbour[];
+    /** The single closest point, or null when the index is empty. */
+    nearest(x: Vec): Neighbour | null;
+    /** The label of an indexed point. */
+    labelOf(index: number): number;
+    /** Class probabilities from similarity-weighted votes (uniform when empty). */
+    proba(x: Vec): number[];
+    predict(x: Float32Array): MlPrediction;
+}
+/** Index `samples` for kNN classification. */
+export declare function fitKnn(samples: readonly MlSample[], opts?: KnnOptions): KnnModel;
+//# sourceMappingURL=knn.d.ts.map

package/dist/ml/classical/knn.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"knn.d.ts","sourceRoot":"","sources":["../../../src/ml/classical/knn.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,KAAK,EAAE,YAAY,EAAE,WAAW,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AACvE,OAAO,EAAmE,KAAK,GAAG,EAAE,MAAM,WAAW,CAAC;AAEtG,MAAM,MAAM,SAAS,GAAG,QAAQ,GAAG,WAAW,CAAC;AAE/C,MAAM,WAAW,UAAU;IACzB,oEAAoE;IACpE,CAAC,CAAC,EAAE,MAAM,CAAC;IACX,MAAM,CAAC,EAAE,SAAS,CAAC;IACnB,yEAAyE;IACzE,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,SAAS;IACxB,KAAK,EAAE,MAAM,CAAC;IACd,2EAA2E;IAC3E,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,qBAAa,QAAS,YAAW,WAAW;IAMxC,QAAQ,CAAC,CAAC,EAAE,MAAM;IAClB,QAAQ,CAAC,MAAM,EAAE,SAAS;IAC1B,QAAQ,CAAC,OAAO,EAAE,MAAM;IAP1B,OAAO,CAAC,QAAQ,CAAC,EAAE,CAAsB;IACzC,OAAO,CAAC,QAAQ,CAAC,EAAE,CAAgB;IACnC,OAAO,CAAC,QAAQ,CAAC,EAAE,CAAgB;gBAGxB,CAAC,EAAE,MAAM,EACT,MAAM,EAAE,SAAS,EACjB,OAAO,EAAE,MAAM;IAK1B,uCAAuC;IACvC,IAAI,IAAI,IAAI,MAAM,CAEjB;IAED,uEAAuE;IACvE,GAAG,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,SAAI,EAAE,MAAM,SAAI,GAAG,IAAI;IAMpC,UAAU,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,EAAE,GAAG,GAAG,MAAM;IAIlC,6CAA6C;IAC7C,UAAU,CAAC,CAAC,EAAE,GAAG,EAAE,CAAC,GAAE,MAAe,GAAG,SAAS,EAAE;IAMnD,iEAAiE;IACjE,OAAO,CAAC,CAAC,EAAE,GAAG,GAAG,SAAS,GAAG,IAAI;IAIjC,qCAAqC;IACrC,OAAO,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM;IAI9B,+EAA+E;IAC/E,KAAK,CAAC,CAAC,EAAE,GAAG,GAAG,MAAM,EAAE;IAWvB,OAAO,CAAC,CAAC,EAAE,YAAY,GAAG,YAAY;CAIvC;AAED,8CAA8C;AAC9C,wBAAgB,MAAM,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,EAAE,IAAI,GAAE,UAAe,GAAG,QAAQ,CAMpF"}

package/dist/ml/classical/knn.js

@@ -0,0 +1,68 @@
+import { classCount, cosine, featureDim, squaredDistance, topProbability } from "./math.js";
+export class KnnModel {
+    k;
+    metric;
+    classes;
+    _x = [];
+    _y = [];
+    _w = [];
+    constructor(k, metric, classes) {
+        this.k = k;
+        this.metric = metric;
+        this.classes = classes;
+        if (!(k >= 1))
+            throw new Error(`KnnModel: k must be ≥ 1 (got ${k})`);
+    }
+    /** How many points the index holds. */
+    get size() {
+        return this._x.length;
+    }
+    /** Add one labelled point (online — the index never needs a refit). */
+    add(x, y = 0, weight = 1) {
+        this._x.push(Float32Array.from(x));
+        this._y.push(y);
+        this._w.push(weight);
+    }
+    similarity(a, b) {
+        return this.metric === "cosine" ? cosine(a, b) : 1 / (1 + Math.sqrt(squaredDistance(a, b)));
+    }
+    /** The `k` closest points, closest first. */
+    neighbours(x, k = this.k) {
+        const all = this._x.map((p, index) => ({ index, similarity: this.similarity(x, p) }));
+        all.sort((a, b) => b.similarity - a.similarity || a.index - b.index);
+        return all.slice(0, Math.min(k, all.length));
+    }
+    /** The single closest point, or null when the index is empty. */
+    nearest(x) {
+        return this.neighbours(x, 1)[0] ?? null;
+    }
+    /** The label of an indexed point. */
+    labelOf(index) {
+        return this._y[index];
+    }
+    /** Class probabilities from similarity-weighted votes (uniform when empty). */
+    proba(x) {
+        const votes = new Array(this.classes).fill(0);
+        for (const n of this.neighbours(x)) {
+            // Shift cosine into (0, 2] so an orthogonal neighbour still casts a small vote.
+            const w = (this.metric === "cosine" ? n.similarity + 1 : n.similarity) * this._w[n.index] + 1e-9;
+            votes[this._y[n.index]] += w;
+        }
+        const sum = votes.reduce((a, b) => a + b, 0);
+        return sum > 0 ? votes.map((v) => v / sum) : votes.map(() => 1 / this.classes);
+    }
+    predict(x) {
+        const scores = this.proba(x);
+        return { scores, confidence: topProbability(scores) };
+    }
+}
+/** Index `samples` for kNN classification. */
+export function fitKnn(samples, opts = {}) {
+    featureDim(samples.map((s) => s.x));
+    const classes = opts.classes ?? classCount(samples.map((s) => s.y));
+    const model = new KnnModel(Math.min(opts.k ?? 5, samples.length), opts.metric ?? "cosine", classes);
+    for (const s of samples)
+        model.add(s.x, s.y, s.weight ?? 1);
+    return model;
+}
+//# sourceMappingURL=knn.js.map

package/dist/ml/classical/knn.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"knn.js","sourceRoot":"","sources":["../../../src/ml/classical/knn.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,UAAU,EAAE,MAAM,EAAE,UAAU,EAAE,eAAe,EAAE,cAAc,EAAY,MAAM,WAAW,CAAC;AAkBtG,MAAM,OAAO,QAAQ;IAMR;IACA;IACA;IAPM,EAAE,GAAmB,EAAE,CAAC;IACxB,EAAE,GAAa,EAAE,CAAC;IAClB,EAAE,GAAa,EAAE,CAAC;IAEnC,YACW,CAAS,EACT,MAAiB,EACjB,OAAe;QAFf,MAAC,GAAD,CAAC,CAAQ;QACT,WAAM,GAAN,MAAM,CAAW;QACjB,YAAO,GAAP,OAAO,CAAQ;QAExB,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;YAAE,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,GAAG,CAAC,CAAC;IACvE,CAAC;IAED,uCAAuC;IACvC,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,EAAE,CAAC,MAAM,CAAC;IACxB,CAAC;IAED,uEAAuE;IACvE,GAAG,CAAC,CAAM,EAAE,CAAC,GAAG,CAAC,EAAE,MAAM,GAAG,CAAC;QAC3B,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC;QACnC,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QAChB,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;IACvB,CAAC;IAED,UAAU,CAAC,CAAM,EAAE,CAAM;QACvB,OAAO,IAAI,CAAC,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,IAAI,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAC9F,CAAC;IAED,6CAA6C;IAC7C,UAAU,CAAC,CAAM,EAAE,IAAY,IAAI,CAAC,CAAC;QACnC,MAAM,GAAG,GAAG,IAAI,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,UAAU,EAAE,IAAI,CAAC,UAAU,CAAC,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;QACtF,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,GAAG,CAAC,CAAC,UAAU,IAAI,CAAC,CAAC,KAAK,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC;QACrE,OAAO,GAAG,CAAC,KAAK,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC;IAC/C,CAAC;IAED,iEAAiE;IACjE,OAAO,CAAC,CAAM;QACZ,OAAO,IAAI,CAAC,UAAU,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,IAAI,CAAC;IAC1C,CAAC;IAED,qCAAqC;IACrC,OAAO,CAAC,KAAa;QACnB,OAAO,IAAI,CAAC,EAAE,CAAC,KAAK,CAAE,CAAC;IACzB,CAAC;IAED,+EAA+E;IAC/E,KAAK,CAAC,CAAM;QACV,MAAM,KAAK,GAAG,IAAI,KAAK,CAAS,IAAI,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC;QACtD,KAAK,MAAM,CAAC,IAAI,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,EAAE,CAAC;YACnC,gFAAgF;YAChF,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,CAAC,UAAU,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAE,GAAG,IAAI,CAAC;YAClG,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAE,CAAE,IAAI,CAAC,CAAC;QACjC,CAAC;QACD,MAAM,GAAG,GAAG,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC;QAC7C,OAAO,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,GAAG,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,CAAC,GAAG,IAAI,CAAC,OAAO,CAAC,CAAC;IACjF,CAAC;IAED,OAAO,CAAC,CAAe;QACrB,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC7B,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,cAAc,CAAC,MAAM,CAAC,EAAE,CAAC;IACxD,CAAC;CACF;AAED,8CAA8C;AAC9C,MAAM,UAAU,MAAM,CAAC,OAA4B,EAAE,OAAmB,EAAE;IACxE,UAAU,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACpC,MAAM,OAAO,GAAG,IAAI,CAAC,OAAO,IAAI,UAAU,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;IACpE,MAAM,KAAK,GAAG,IAAI,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,EAAE,OAAO,CAAC,MAAM,CAAC,EAAE,IAAI,CAAC,MAAM,IAAI,QAAQ,EAAE,OAAO,CAAC,CAAC;IACpG,KAAK,MAAM,CAAC,IAAI,OAAO;QAAE,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,MAAM,IAAI,CAAC,CAAC,CAAC;IAC5D,OAAO,KAAK,CAAC;AACf,CAAC"}