Weights Decaying about bindsnet

[djsaunde]

I think it would be best to implement this in the abstract base class LearningRule, as part of its update() function. Then, any class that subclasses LearningRule would call the LearningRule's update() function as part of its update(), where weight decay would be implemented.

We also need to decided what the form of the weight decay is. Is it something like:

w -= delta # for delta small

or:

w -= delta * w # for delta small

or something else? How is it implemented in Chetan's code? On another note, we should create a private GitHub repository with Chetan's code so we can all access it.

[Hananel-Hazan]

Agreed

[djsaunde]

Has Chetan's code been published in a private repo? If so, could you invite me to collaborate on it?