Scaling Equilibrium Propagation to Deeper Neural Network Architectures