Conceptual

Feedforward Neural Networks

Layers of linear transforms and nonlinearities with no cycles: information flows input to output in one pass. The building block RNNs extend by adding a loop over time.