Overview

Research and experiments building up to an implementation of a reinforcement learning trader based on work by Moody and Saffell [1]

Notebooks

• 1_linreg_mse_grad_descent - DIY gradient descent for MSE of regression.
• 2_sympy_partial_derivatives - Symbolicly derive partial derivatives for MSE using Python Sympy library.
• 3.1_reinforcement_trading_offline_simple - Implementation of offline trader from [1]. Learns an ARIMA style model using a DIY 1 layer neural network and tanh activation.
• 3.1_reinforcement_trading_offline_simple_derivs - Derivatives for above notebook.
• 3.2_reinforcement_trading_offline_pytorch - Implementation of offline trader from [1]. Uses PyTorch and LSTM.

Next steps

• Add online learning implementation implementation using moving averages
• Add train/test/validate split to notebooks

References

[1] J. Moody and M. Saffell, "Learning to trade via direct reinforcement," in IEEE Transactions on Neural Networks, vol. 12, no. 4, pp. 875-889, July 2001, doi: 10.1109/72.935097.