Replace Dense layer with LSTM layer: input shape

AI & Machine learning
1

Course Name: Deep Reinforcement Learning in Trading, Section No: 15, Unit No: 7, Unit type: Notebook

To use as NN one LSTM we need to give as input something with the following shape (samples, timesteps, features) but in your NN the input to Dense layer is (138,none), could you explain how to get the correct input shape for one LSTM to replace Dense layer?

Thank you

Guido

2

Hi Guido,



As you rightly said, a 3-D input os required for the LSTM layer, you need to keep the shape as (samples, timesteps, features).



For e.g., if your RL model has a lookback of 30 days, and you are having 2000 days of historical data, and the number of features the model being trained on is 13, then the input of LSTM layer would be of the shape (1970,30,13).



Hope this helps!