Contributions to RatInABox are welcome. Pull request them into the here.
If you wish to add a new Neurons subclass you should follow the instructions found within the header comments of the Neurons class (code here). N.b. your new Neurons subclass can be a direct subclass of Neurons itself (e.g. MyNeuron(Neuron)) or it can be a subclass of one of the other Neurons subclasses (e.g. MyFancyPlaceCells(PlaceCells) or MyFancyFeedForwardLayerWithItsOwnFancyLearningRule(FeedForwardLayer)).
We provide two example contrib files for you to refer to. These files can also be run standalone to show mini-demos.:
- PhasePrecessingPlaceCells.py: Place cells (1 or 2D) which also phase precess.
- ValueNeuron.py: A neuron which uses TD learning to approximate a value function as a sum of inpouts from a downstream layer. Is a subclass of
FeedForwardLayer - PlaneWaveNeurons.py: A layer of neurons. Each neurons rate map is a plane wave with a random orientation and phase offset.
Once you have made a contribution (myContribution.py), reinstall RatInABox and import using
import ratinabox
from ratinabox.contribs.myContribution import *