This is my implementation of Hinton's forward-forward algorithm for training neural networks. I've primarily focused on the recurrent structure presented in the GLOM paper, since I find it the most probable.
- Implement local goodness scores
- How would the brain know how to reverse the gradient direction in sleep?