/GRU

A C++ implementation of a 3 layer Gated Recurrent Unit (GRU) using no libraries other than Eigen for Matrices.

Primary LanguageC++

This is an implementation of a GRU in C++ using no libraries other than Eigen for Matrices. This is a barebones implementation that uncovers the complexity of backpropagation. All equations have been derived and implemented.

Create a folder for weights called Weights

Usage -

make
./train
./test `epoch_number` `loss`

e.g. ./test 66 0.397798