ts-classify 0.1.0 → 0.1.1

package/LICENSE.md

@@ -0,0 +1,20 @@
+Copyright (c) 2026 Bart Riepe
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
+LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
+WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

package/README.md

@@ -1,123 +1,31 @@
-# SVM
+# ts-classify
 
-A fast Support Vector Machine implementation in Rust, targeting both native and WebAssembly.
+Fast text classification via SVM, compiled to WebAssembly.
 
 ## Features
 
-- **Binary SVM** with two solvers:
-  - **Coordinate Descent** (default) - O(N) per pass, fast for large datasets
-  - **SMO** - O(N²), useful for kernel SVMs (future)
-- **Multiclass classification**:
-  - **One-vs-Rest (OvR)** - K classifiers for K classes, fast prediction
-  - **One-vs-One (OvO)** - K(K-1)/2 classifiers, often more accurate
+- **Multiclass classification** via One-vs-Rest (OvR) or One-vs-One (OvO)
+- **Binary SVM** with coordinate descent solver (O(N) per pass)
 - **Sparse vectors** - efficient for high-dimensional data (text, trigrams)
-- **Parallel training** via rayon (native only)
-- **Binary serialization** via bincode
-- **WebAssembly support** with JS-friendly API
+- **Binary serialization** - train once, load instantly
+- **Written in Rust**, compiled to WASM for near-native performance in the browser
 
 ## Installation
 
-```toml
-[dependencies]
-svm = { path = "." }
-```
-
-## Usage
-
-### Binary SVM
-
-```rust
-use svm::{SVM, SparseVec};
-
-// Sparse format: Vec<(index, value)>
-let samples: Vec<SparseVec> = vec![
-    vec![(0, -2.0), (1, 0.0)],
-    vec![(0, -1.0), (1, 1.0)],
-    vec![(0, 1.0), (1, 0.0)],
-    vec![(0, 2.0), (1, 1.0)],
-];
-let labels = vec![-1.0, -1.0, 1.0, 1.0];
-
-let mut svm = SVM::new();
-svm.train_sparse(&samples, &labels);
-
-assert_eq!(svm.predict_sparse(&[(0, -3.0)]), -1.0);
-assert_eq!(svm.predict_sparse(&[(0, 3.0)]), 1.0);
-```
-
-### Multiclass (One-vs-Rest)
-
-```rust
-use svm::{OneVsRestSVM, SparseVec};
-
-let samples: Vec<SparseVec> = vec![
-    vec![(0, -2.0), (1, -2.0)],  // class 0
-    vec![(0, 2.0), (1, -2.0)],   // class 1
-    vec![(0, 0.0), (1, 2.0)],    // class 2
-    // ... more samples
-];
-let labels = vec![0, 1, 2, /* ... */];
-
-let mut svm = OneVsRestSVM::new();
-svm.train_sparse(&samples, &labels);
-
-let predicted_class = svm.predict_sparse(&[(0, 0.0), (1, 3.0)]);
-```
-
-### Custom Configuration
-
-```rust
-use svm::{SVM, TrainConfig, Solver};
-
-let config = TrainConfig::coordinate_descent()
-    .with_c(0.5)           // regularization (default: 1.0)
-    .with_max_iter(500)    // max iterations (default: 1000)
-    .with_tol(0.001);      // tolerance (default: 0.01)
-
-let mut svm = SVM::new();
-svm.train_sparse_with_config(&samples, &labels, &config);
-
-// Or use SMO solver
-let config = TrainConfig::smo()
-    .with_c(1.0)
-    .with_max_iter(10000);
-```
-
-### Serialization
-
-```rust
-// Save
-let bytes = svm.to_bytes()?;
-std::fs::write("model.bin", &bytes)?;
-
-// Load
-let bytes = std::fs::read("model.bin")?;
-let svm = SVM::from_bytes(&bytes)?;
-```
-
-## WebAssembly
-
-Build for WASM (without parallel feature):
-
 ```bash
-cargo build --release --target wasm32-unknown-unknown --no-default-features
+npm install ts-classify
 ```
 
-Or use wasm-pack:
-
-```bash
-wasm-pack build --target web --no-default-features
-```
-
-### JavaScript API
+## Usage
 
 ```javascript
-import { JsOneVsRestSVM } from 'svm';
+import init, { JsOneVsRestSVM } from 'ts-classify';
+
+await init();
 
 const svm = new JsOneVsRestSVM();
 
 // Flat sparse format: [idx, val, idx, val, ...]
-const sampleFlat = new Float64Array([0, -2.0, 1, 0.0]);  // single sample
 const samplesFlat = new Float64Array([
     0, -2.0, 1, 0.0,  // sample 0: 2 pairs
     0, 1.0, 1, 1.0,   // sample 1: 2 pairs
@@ -127,7 +35,7 @@ const labels = new Int32Array([0, 1]);
 
 svm.train(samplesFlat, sampleLengths, labels);
 
-const prediction = svm.predict(sampleFlat);
+const prediction = svm.predict(new Float64Array([0, -2.0, 1, 0.0]));
 
 // Serialize/deserialize
 const bytes = svm.to_bytes();
@@ -136,13 +44,6 @@ const loaded = JsOneVsRestSVM.from_bytes(bytes);
 
 ## Performance
 
-Coordinate descent vs SMO on 50-class OvR:
-
-| Solver | Train Time |
-|--------|------------|
-| Coordinate Descent | 6ms |
-| SMO | 29s |
-
 Scaling (coordinate descent, OvR):
 
 | Classes | Classifiers | Train | Predict |
@@ -152,16 +53,18 @@ Scaling (coordinate descent, OvR):
 | 100 | 100 | 1.4ms | 1.6µs |
 | 5000 | 5000 | ~50-100ms | ~80µs |
 
-## Feature Flags
+## Building from source
 
-- `parallel` (default) - Enable parallel training via rayon. Disable for WASM.
+Requires [Rust](https://rustup.rs/) and [wasm-pack](https://rustwasm.github.io/wasm-pack/installer/).
 
-```toml
-# Native with parallel
-svm = { path = "." }
+```bash
+wasm-pack build --target web --no-default-features
+```
 
-# WASM (no parallel)
-svm = { path = ".", default-features = false }
+Run tests (native):
+
+```bash
+cargo test --release
 ```
 
 ## License

package/package.json

@@ -2,7 +2,7 @@
   "name": "ts-classify",
   "type": "module",
   "description": "Fast text classification with SVM and Nearest Centroid (WebAssembly)",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "license": "MIT",
   "repository": {
     "type": "git",