@wlearn/liblinear

LIBLINEAR v2.50 compiled to WebAssembly. Linear classification and regression in browsers and Node.js.

Based on LIBLINEAR v2.50 (BSD-3-Clause). Zero dependencies. ESM.

Install

npm install @wlearn/liblinear

Quick start

import { LinearModel } from '@wlearn/liblinear'

const model = await LinearModel.create({
  solver: 'L2R_LR',
  C: 1.0
})

// Train -- accepts number[][] or { data: Float64Array, rows, cols }
model.fit(
  [[1, 2], [3, 4], [5, 6], [7, 8]],
  [0, 0, 1, 1]
)

// Predict
const preds = model.predict([[2, 3], [6, 7]])  // Float64Array

// Probabilities (logistic regression solvers only)
const probs = model.predictProba([[2, 3], [6, 7]])  // Float64Array (nrow * nclass)

// Score
const accuracy = model.score([[2, 3], [6, 7]], [0, 1])

// Save / load
const buf = model.save()  // Uint8Array
const model2 = await LinearModel.load(buf)

// Clean up -- required, WASM memory is not garbage collected
model.dispose()
model2.dispose()

Typed matrix input (fast path)

For performance-critical code and AutoML integration, pass typed matrices directly:

const X = {
  data: new Float64Array([1, 2, 3, 4, 5, 6, 7, 8]),
  rows: 4,
  cols: 2
}
const y = new Float64Array([0, 0, 1, 1])

model.fit(X, y)

This avoids the conversion cost of nested arrays.

Input coercion policy

Control how non-typed inputs are handled:

// Default: convert silently
const model = await LinearModel.create({ coerce: 'auto' })

// Warn once per instance when conversion happens
const model = await LinearModel.create({ coerce: 'warn' })

// Throw if input is not already a typed matrix
const model = await LinearModel.create({ coerce: 'error' })

API

LinearModel.create(params?)

Async factory. Loads WASM module, returns a ready-to-use model.

Parameters:

• solver -- solver type string or number (default: 'L2R_LR')
• C -- regularization parameter (default: 1.0)
• eps -- stopping tolerance (default: 0.01)
• bias -- bias term, < 0 disables (default: -1)
• p -- epsilon in SVR loss (default: 0.1)
• coerce -- input coercion: 'auto' | 'warn' | 'error' (default: 'auto')

model.fit(X, y)

Train on data. Returns this for chaining.

• X -- number[][] or { data: Float64Array, rows, cols }
• y -- number[] or Float64Array

model.predict(X)

Returns Float64Array of predicted labels.

model.predictProba(X)

Returns Float64Array of shape nrow * nclass (row-major probabilities). Only available for logistic regression solvers (L2R_LR, L1R_LR, L2R_LR_DUAL).

model.decisionFunction(X)

Returns Float64Array of decision values.

model.score(X, y)

Returns accuracy (classification) or R-squared (regression).

model.save()

Returns Uint8Array (native LIBLINEAR model format).

LinearModel.load(buffer)

Loads from Uint8Array. Returns Promise<LinearModel>.

model.dispose()

Free WASM memory. Required. Idempotent.

model.getParams() / model.setParams(p)

Get/set hyperparameters. Enables AutoML grid search and cloning.

LinearModel.defaultSearchSpace()

Returns default hyperparameter search space for AutoML.

Solver types

| Name | Code | Task | |------|------|------| | L2R_LR | 0 | L2-regularized logistic regression | | L2R_L2LOSS_SVC_DUAL | 1 | L2-loss SVM (dual) | | L2R_L2LOSS_SVC | 2 | L2-loss SVM (primal) | | L2R_L1LOSS_SVC_DUAL | 3 | L1-loss SVM (dual) | | MCSVM_CS | 4 | Multi-class SVM (Crammer-Singer) | | L1R_L2LOSS_SVC | 5 | L1-regularized L2-loss SVM | | L1R_LR | 6 | L1-regularized logistic regression | | L2R_LR_DUAL | 7 | L2-regularized logistic regression (dual) | | L2R_L2LOSS_SVR | 11 | L2-loss SVR (primal) | | L2R_L2LOSS_SVR_DUAL | 12 | L2-loss SVR (dual) | | L2R_L1LOSS_SVR_DUAL | 13 | L1-loss SVR (dual) |

Resource management

WASM heap memory is not garbage collected. Call .dispose() on every model when done. A FinalizationRegistry safety net warns if you forget, but do not rely on it.

Build from source

Requires Emscripten (emsdk) activated.

git clone --recurse-submodules https://github.com/wlearn-org/liblinear-wasm
cd liblinear-wasm
npm run build
npm test

If you already cloned without --recurse-submodules:

git submodule update --init

License

BSD-3-Clause (same as upstream LIBLINEAR)