ml-pls

Partial Least Squares (PLS), Kernel-based Orthogonal Projections to Latent Structures (K-OPLS) and NIPALS based OPLS

PLS regression algorithm based on the Yi Cao implementation:

PLS Matlab code

K-OPLS regression algorithm based on this paper.

K-OPLS Matlab code

OPLS implementation based on the R package Metabomate using NIPALS factorization loop.

Installation

$ npm i ml-pls

Usage

PLS

import { PLS } from 'ml-pls';

const X = [
  [0.1, 0.02],
  [0.25, 1.01],
  [0.95, 0.01],
  [1.01, 0.96],
];
const Y = [
  [1, 0],
  [1, 0],
  [1, 0],
  [0, 1],
];
const options = {
  latentVectors: 10,
  tolerance: 1e-4,
};

const pls = new PLS(options);
pls.train(X, Y);

OPLS-R

import {
  getNumbers,
  getClassesAsNumber,
  getCrossValidationSets,
} from 'ml-dataset-iris';
import { OPLS } from 'ml-pls';

const cvFolds = getCrossValidationSets(7, { idx: 0, by: 'trainTest' });
const data = getNumbers();
const irisLabels = getClassesAsNumber();

const model = new OPLS(data, irisLabels, { cvFolds });
console.log(model.mode); // 'regression'

The OPLS class is intended for exploratory modeling, that is not for the creation of predictors. Therefore there is a built-in k-fold cross-validation loop and Q2y is an average over the folds. Q2y is an array with one entry per fitted component:

console.log(model.model[0].Q2y[0]);

should give 0.9209227614652846

OPLS-DA

import {
  getNumbers,
  getClasses,
  getCrossValidationSets,
} from 'ml-dataset-iris';
import { OPLS } from 'ml-pls';

const cvFolds = getCrossValidationSets(7, { idx: 0, by: 'trainTest' });
const data = getNumbers();
const irisLabels = getClasses();

const model = new OPLS(data, irisLabels, { cvFolds });
console.log(model.mode); // 'discriminantAnalysis'
// auc is an array with one entry per fitted component
console.log(model.model[0].auc[0]); // 0.6816

If for some reason a predictor is necessary the following code may serve as an example

Prediction

import {
  getNumbers,
  getClassesAsNumber,
  getCrossValidationSets,
} from 'ml-dataset-iris';
import { OPLS } from 'ml-pls';

// get frozen folds for testing purposes
const { testIndex, trainIndex } = getCrossValidationSets(7, {
  idx: 0,
  by: 'trainTest',
})[0];

// Getting the data of selected fold
const irisNumbers = getNumbers();
const testData = irisNumbers.filter((el, idx) => testIndex.includes(idx));
const trainingData = irisNumbers.filter((el, idx) => trainIndex.includes(idx));

// Getting the labels of selected fold
const irisLabels = getClassesAsNumber();
const testLabels = irisLabels.filter((el, idx) => testIndex.includes(idx));
const trainingLabels = irisLabels.filter((el, idx) => trainIndex.includes(idx));

const model = new OPLS(trainingData, trainingLabels);
console.log(model.mode); // 'regression'
const prediction = model.predict(testData, { trueLabels: testLabels });
// Get the predicted Q2 value
console.log(prediction.Q2y); // 0.9243354801393767

K-OPLS

import Kernel from 'ml-kernel';
import { KOPLS } from 'ml-pls';

const kernel = new Kernel('gaussian', {
  sigma: 25,
});

const X = [
  [0.1, 0.02],
  [0.25, 1.01],
  [0.95, 0.01],
  [1.01, 0.96],
];
const Y = [
  [1, 0],
  [1, 0],
  [1, 0],
  [0, 1],
];

const cls = new KOPLS({
  orthogonalComponents: 10,
  predictiveComponents: 1,
  kernel: kernel,
});

cls.train(X, Y);

const {
  prediction, // prediction
  predScoreMat, // Score matrix over prediction
  predYOrthVectors, // Y-Orthogonal vectors over prediction
} = cls.predict(X);

console.log(prediction);
console.log(predScoreMat);
console.log(predYOrthVectors);

API Documentation

License

MIT