Implemented simple 2-d data fitter with p5.js and convnet.js libraries, to play around with stochastic gradient descent fitting of simple data.

In this demo, the user can enter a series of points in the (X, Y) coordinate space, and the neural network will try to come up with a Y = f(X) function to fit the data.

It’s cool to see how these things actually try to fit the data interactively, and by choosing different layers, such as rectifier layers, sigmoid layers, how they behave differently.  Rectifier layers tend to be more ‘brain’ like, as they are on/off, while sigmoid layers tend to be more of a smoothing function.  When they are used together in a multi-layer network, things get a bit interesting

Try demo here.