dfm / optimize.js

A tool for robust numerical optimization running right in your browser

optimize.js

This kind of optimization. Not that.

This was ported from the scipy.optimize and we seem to be about an order of magnitude faster than the Python version but that hasn't been too well tested.

Usage

To optimize some dumb function, just run

my_function = function (x) {
    return optimize.vector.dot(x, x);
};
xopt = optimize.fmin(my_function, [5.0, -3.4, 1.7, 16.3, 0.17]);

And this should say something like:

Converged in 349 iterations.
Function value = 3.954810202072493e-7

And xopt should end up being something like

[-0.0003993727670733724, -0.00027257793115893254, -0.0003811443958917447, 0.00012560283306980614, 0.00002523018782488204]

You can also include some options as follows:

xopt = optimize.fmin(my_function, [5.0, -3.4, 1.7, 16.3, 0.17], {ftol: 1e-7, maxiter: 1000});

To Do

• Write tests.
• Implement some better algorithms.
• Make some demos.