Recurrent Reinforcement Learning Implementation using Matlab/Octave

20210321 Update: A PyTorch-port of this repo is available at ceshine/RRL_PyTorch.

Reference: Stock Trading with Recurrent Reinforcement Learning (RRL) By Gabriel Molina

File Description

Core Functions

• costFunction.m
• updateFt.m
• rewardFunction.m
• featureNormalize.m
• sharpRatio.m

Utility Functions

• checkRRLGradient.m : Verify the correctness of gradient function in cost function
• getNumericalGradient.m : Approximate gradient (inefficiently) for verification

Test Function:

• testTWSE.m : Use Taiwan Weighted Stock Index from Taiwan Stock Exchange.
• testDAX.m : Use DAX index of Germany.