@ai-on-browser/data-analysis-models/lib/model/pls.js

Data analysis model package without any dependencies

import Matrix from '../util/matrix.js'

/**
 * Partial least squares regression
 */
export default class PLS {
	/**
	 * @param {number} l Limit on the number of latent factors
	 */
	constructor(l) {
		this._l = l
	}

	/**
	 * Initialize model.
	 * @param {Array<Array<number>>} x Training data
	 * @param {Array<Array<number>>} y Target values
	 */
	init(x, y) {
		this._x = Matrix.fromArray(x)
		this._y = Matrix.fromArray(y)
	}

	/**
	 * Fit model.
	 */
	fit() {
		if (this._y.cols === 1) {
			;[this._b, this._b0] = this._pls1()
		} else {
			throw ''
		}
	}

	_pls1() {
		// https://ja.wikipedia.org/wiki/部分的最小二乗回帰
		let x = this._x.copy()
		let w = x.tDot(this._y)
		w.div(w.norm())

		const ws = []
		const ps = []
		const qs = []
		for (let k = 0; k < this._l; k++) {
			const t = x.dot(w)
			const tk = t.tDot(t).toScaler()
			t.div(tk)
			const p = x.tDot(t)
			const qk = this._y.tDot(t).toScaler()

			ps.push(p.value)
			qs.push(qk)
			ws.push(w.value)
			if (qk === 0) break

			const xsub = t.dot(p.t)
			xsub.mult(tk)
			x.sub(xsub)
			w = x.tDot(this._y)
		}
		const W = Matrix.fromArray(ws).t
		const P = Matrix.fromArray(ps).t
		const q = new Matrix(qs.length, 1, qs)

		const B = W.dot(P.tDot(W).solve(q))
		const B0 = qs[0] - P.col(0).tDot(B).toScaler()
		return [B, B0]
	}

	/**
	 * Returns predicted values.
	 * @param {Array<Array<number>>} x Sample data
	 * @returns {Array<Array<number>>} Predicted values
	 */
	predict(x) {
		x = Matrix.fromArray(x)
		return Matrix.add(x.dot(this._b), this._b0).toArray()
	}
}