Linear Regression Plotter

Purpose:

Inspired by the Andrew Ng's Machine Learning online course, I made my own linear regression plotter in vanilla JavaScript in ECMA6. The library mathjs was added but hasn't been used. Suffice it to say I learned plenty through this project of how JavaScript works and how mathematics works.

Next Steps:

If continued, the project will add support especially contour plots, and gradient descent, and ancillarily any dataset scale, n dimensions of data, and selecting dimension shown.

Setup:

Download and just run.

A Taste Of The Code:

Defined in these Coursera notes, the mathematical cost function J here:

is adapted to the javascript function EvaluateY below:

    /**
     * Evaluate y over all dimensions of x.  e.g. y = x_0 * θ_0 + x_1 * θ_1 + ... + x_n * θ_n
     *
     * @param xs Array of x values
     * @returns {number} Value of y on line at given xs(x_0, x_1, x_2 ... x_n)
     */
    this.EvaluateY = (xs) => {

        let smallerArrayLength = (model.thetas.length < xs.length) ? model.thetas.length : xs.length;

        let y = 0;

        for (let i = 0; i < smallerArrayLength; i++) {
            let x = xs[i];
            let θ = model.thetas[i];
            y += x * θ;
        }

        return y;
    };

Which can begin a detailed conversation of differences applying the math-world to the js-world.