Implementation of some Neural Network Architectures and training methods in PyTorch, including Autoencoder (AE), FeedForward (FFNN), Recursive (RNN) and Extreme Learning Machine (ELM). FNN are trained in batch or online with the Extended Kalman Filter. The ELM are trained with least squared or online with recursive least squares. RNN is trained with Truncated Backprobagation trough Time.