Personae - RL & SL Methods and Envs For Quantitative Trading
Personae is a repo that implements papers proposed methods in Deep Reinforcement Learning & Supervised Learning and applies them to simulate future financial trends.
Based on Ceruleanacg's repo Personae's framework, I've personalized it to support TD's mutual funds, which I am currently investing in. Currently retraining it to more accurately predict my funds of interest. As of July 10th, 2018, I've earned $378 off mutual funds! :)
Implemented Models (by Ceruleanacg)
Deep Deterministic Policy Gradient (DDPG)
Implement of DDPG with TensorFlow.
arXiv:1509.02971: Continuous control with deep reinforcement learning
Implement of Double-DQN with TensorFlow.
arXiv:1509.06461: Deep Reinforcement Learning with Double Q-learning
Implement of Dueling-DQN with TensorFlow.
arXiv:1511.06581: Dueling Network Architectures for Deep Reinforcement Learning
Implement of Policy Gradient with TensorFlow.
NIPS. Vol. 99. 1999: Policy gradient methods for reinforcement learning with function approximation
Implement of arXiv:1704.02971, DA-RNN with TensorFlow.
Implement of TreNet with TensorFlow.
Implement of simple LSTM based model with TensorFlow.
arXiv:1506.02078: Visualizing and Understanding Recurrent Networks
Before you start testing, following requirements are needed.
The original repo had a GPU docker, but I've included a CPU dockerfile as well.