This repo is the implementation of perturbation based algorithms for training neural networks - we evaluate the scalability of node perturbation with network width and depth on modern datasets such as MNIST, CIFAR, etc.
