Weightless Neural Nets

WNNs are here tested on character images. You can select the tuple size as 4 or 8, and allow linear/random tuple mapping.

12 different characters are shown - use the mouse to click on one to select it, so it becomes the input to the network.

The RAM neurons are in the middle: you can see each location as 0 or 1.

When you press Learn, tuples are found from the selected input and taught in each RAM - certain RAM locations are now 1.

When you press Analyse, the input is sampled in the same way and the number of RAM locations found with a 1 is shown.

Analyse All causes each character to be tested and you see how many neurons 'recognise' each character.

You can restart by clearing the RAMs to 0.

In Demo Mode, you see the tuples being formed and specific RAM locations accessed.

Tuple Size: 4; 8     Demo Mode :

Tuple Mapping: Linear; Random;