Factor analysis

Algorithm overview

Factor analysis explains correlated observations with a smaller set of latent variables while estimating separate noise variance for every feature. It is useful when measurement noise matters more than PCA's total-variance objective.

JavaScript implementation

FactorAnalysis runs the sklearn-style EM/SVD update in JavaScript and supports optional varimax or quartimax rotation.

Quick start

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

const X = [[1, 2, 1], [2, 4, 1], [3, 6, 2], [4, 8, 2]];
const model = new Decomposition.FactorAnalysis({ nComponents: 2 });
const scores = model.fitTransform(X);
console.log(scores);

Detailed API reference

Constructor options are nComponents, tol, maxIter, noiseVarianceInit, and rotation. Fitted state includes components, mean, noiseVariance, loglike, and nIter, plus getCovariance() and getPrecision().