This repo contains the code of the original equilibrium propagation (EP) paper. The code is written in Theano, the framework once developed by Mila.
For more recent code:
- This repo contains code written in PyTorch.
- This repo contains code for 'holomorphic equilibrium propagation', written in Jax.
Other repositories with code for EP: 1, 2, 3.
- Download the code from GitHub:
git clone https://github.com/bscellier/Towards-a-Biologically-Plausible-Backprop
cd Towards-a-Biologically-Plausible-Backprop- To train a Hopfield network (with 1 hidden layer by default) with Eqprop, run the command:
THEANO_FLAGS="floatX=float32, gcc.cxxflags='-march=core2'" python train_model.py- Once the network is trained, use the GUI by running the command:
THEANO_FLAGS="floatX=float32, gcc.cxxflags='-march=core2'" python gui.py net1