This repository contains the PyTorch implementation for the post-synaptic potential (PSP) regularization technique. All the differentiation is performed by autograd: the post-synaptic potentials are passed along with the overall output of the network, during the forward propagation step.
The following models are included:
- LeNet-5 for MNIST and Fashion-MNIST
- ResNet-18 for CIFAR-10
- MobileNetv2 for CIFAR-10
- All-CNN-C for CIFAR-10