Implementing an ADMM based optimization approach as an alternative to backpropagation for training neural networks.