Deep Learning Algorithm in Catalyst

Using deep learning and reinforcement learning in a Enigma Catalyst algorithm.

In this algorithm, implemented in Enigma Catalyst, is used Deep Deterministic Policy Gradient for Portfolio Management. For the actor-critic agent is used a Convolutional Neural Network architecture based in Jiang, Xyu and Liang work.

The algorithm is implemented in Enigma Catalyst using Poloniex Exchange data. Catalyst is based in Zipline which is the base code for Quantopian platform. This make easy extrapolate the algorithm for these recognized platforms.

Before executing the algorithm you will need to install Catalyst in first place. You can use the Catalyst installation instruccions.

References:

• Practical Deep Reinforcement Learning Approach for Stock Trading, (Zhuoran Xiong, Xiao-Yang Liu, Shan Zhong, Hongyang Yang, Anwar Walid)
• Deep Reinforcement Learning for Pair Trading Using Actor Critic, (Yichen Shen, Yiding Zhao)
• A Deep Reinforcement Learning Framework for the Finantial Portfolio Management Problem, (Zhengyao Jiang, Dixing Xyu, Jinjun Liang)
• [Deep Deterministic Policy Gradient in TensorFlow] (https://pemami4911.github.io/blog/2016/08/21/ddpg-rl.html), (Patrick Emami)