hnn

Documentation Status

Hebbian Neural Network

Features

  • Everything

Some ideas we were talking about:

  1. Dopamine is globally applied and strengthens Hebbian connections that have been recently potentiated
  2. Neurons require a baseline firing rate to create tenuous Hebbian connections that can be reenforced when the right network forms
  3. If a connection is 0 weighted, hebbian learning cannot take place (this allows genes to provide structure)
  4. Not connecting is more powerful than connecting. This provides regularization.

Credits