The simulation below shows a toy binary problem with a few data points of class 0 (red) and 1 (green). The network is set up as:
Feel free to change this, the text area above gets eval()'d when you hit the button and the network gets reloaded. Every 10th of a second, all points are fed to the network multiple times through the trainer class to train the network. The resulting predictions of the network are then "painted" under the data points to show you the generalization.
On the right we visualize the transformed representation of all grid points in the original space and the data, for a given layer and only for 2 neurons at a time. The number in the bracket shows the total number of neurons at that level of representation. If the number is more than 2, you will only see the two visualized but you can cycle through all of them with the cycle button.
CLICK: Add red data point
SHIFT+CLICK: Add green data point
ALT+CLICK: Remove closest data point