lanceshih / trading-brain

Framework for deep learning in Trading-Gym environment

Trading Brain is a framework example for implementing and testing trading strategies. It is composed of mainly three components communicating through APIs:

• Brain
• Memory
• Agent

This library can be used to test agents with the Trading-Gym.

Installation

Install packages in requirements.txt file

Roll out your own Agent

To create your own agent, it must inherit from the Agent base class which can be found at 'tbrn/base/agent.py'. It consists of three basic methods that need to be overridden in order to implement your own logic:

• act: returns the action chosen by the agent.
• observe: returns a real value (can be the loss in the case of a KerasAgent for instance). This method is where the learning logic of the agent is located. Can be blank for dummy agents.
• end: any logic at the end of an episode.

Examples

One example can be found in examples/

• Simple keras agent (examples/keras_example.py)
• Dueling Double DQN tensorflow agent (examples/tf_example.py)

Read more about this example at our Trading Gym

Copyright © 2017 RKR Epsilon UK Ltd. All rights reserved.