Hw 9.3.2

9.3.2). Train the model on the first 4 years of data using plain RNN, GRU and LSTM cells (for a single layer). How well can the model

In the remaining questions after this, we are instructed to evaluate using “the model”. Is this the best model we have from 9.3.2 or all 3 of them?

Ideally, you should train each of them separately and compare with each other for all 3 of them.

While it may take you a long time to train them again, so you can pick the best architecture within RNN, GRU and LSTM, and just simply compare the performance on 3.3.