Feedback-Alignment

Implementations of Random Feedback Alignment (RFA), DFA, and variants.

Running

python3 main.py --lr=0.05 --bs=128 --epochs=100 --flow=autodiff

lr Learning rate
bs Batch size
epochs Number of epochs to train for
flow The gradient flow scheme used. Possible values:
- autodiff Backprop with Tensorflow's built-in reverse mode automatic differention
- rfa Random feedback alignment
- dfa Direct feedback alignment