Label Spreading in JavaScript with @kanaries/ml

Learn what Label Spreading does, when to use it, and how to run LabelSpreading in JavaScript or TypeScript with @kanaries/ml for browser and Node.js applications.

Algorithm overview

LabelSpreading performs smoother semi-supervised label diffusion with regularization to reduce over-confident propagation.

This algorithm is especially useful when:

You want semi-supervised learning with stronger stability than pure propagation.
Graph neighborhoods are useful but somewhat noisy.
You need better robustness for low-label datasets.

@kanaries/ml provides Label Spreading in JavaScript for semi-supervised problems where you want a smoother, regularized alternative to raw label propagation. That is useful for annotation tools and partially labeled datasets where graph-based learning should be robust to noisy neighborhoods.

Because the implementation runs in JS, teams can connect interactive labeling workflows, similarity features, and propagation-based learning without a separate Python service layer.

Quick start

LabelSpreading in Python vs JavaScript / TypeScript

If you searched for "LabelSpreading in JavaScript" or "LabelSpreading in TypeScript", this section maps the familiar scikit-learn call to the equivalent @kanaries/ml usage for browser and Node.js runtimes.

Python

scikit-learn

from sklearn.semi_supervised import LabelSpreading

X = [[0, 0], [0.1, 0.2], [1, 1], [1.1, 0.9]]
y = [0, -1, 1, -1]

model = LabelSpreading(alpha=0.2)
model.fit(X, y)
pred = model.predict(X)

JavaScript / TypeScript

@kanaries/ml

import { SemiSupervised } from '@kanaries/ml';

const X = [[0, 0], [0.1, 0.2], [1, 1], [1.1, 0.9]];
const y = [0, -1, 1, -1];

const model = new SemiSupervised.LabelSpreading({ alpha: 0.2 });
model.fit(X, y);
const pred = model.predict(X);

Quick JavaScript example

import { SemiSupervised } from '@kanaries/ml';

const X = [[0, 0], [0.1, 0.2], [1, 1], [1.1, 0.9]];
const y = [0, -1, 1, -1];

const model = new SemiSupervised.LabelSpreading({ alpha: 0.2 });
model.fit(X, y);
const pred = model.predict(X);

Detailed API reference

interface LabelSpreadingOptions {
    kernel?: 'rbf' | 'knn' | ((X: number[][], Y: number[][]) => number[][]);
    gamma?: number;
    nNeighbors?: number;
    alpha?: number;
    maxIter?: number;
    tol?: number;
}
constructor(options: LabelSpreadingOptions = {})

Label spreading assigns labels to unlabeled data using a normalized graph and soft clamping controlled by alpha.

Methods

fit(trainX: number[][], trainY: number[]): void
predict(testX: number[][]): number[]
predictProba(testX: number[][]): number[][]

Implementation workflow

Create labeled/unlabeled split and construct feature graph inputs.
Fit LabelSpreading and monitor convergence behavior.
Tune kernel and regularization parameters with validation labels.

JavaScript deployment notes

Prefer Label Spreading when graph noise or over-confident propagation is a concern.
Tune kernel and regularization settings against a validation subset of known labels.
Use it to improve label coverage before training a final supervised model rather than as the only production model.

Label Spreading