bgalbraith / somnia

Self-Organizing Maps as Neural Interactive Art

SOMINA: Self-Organizing Maps as Neural Interactive Art

How to use:

python somnia.py --source <image>

There are a variety of other command line arguments you can pass in to set the initial properties of the network.

This will launch a window and the model will begin to train. To save a copy of the state of the network, press the space bar. You can change the neighborhood function properties in the following ways:

• window function - constant (1), linear (2), Gaussian (3), difference of Gaussians (4)
• radius - increase (right), decrease (left)
• learning rate - increase (up), decrease (down)
• tiling effect - increase (right brace), decrease (left brace)