Status: Development (expect API changes and major updates)

Spectrum-Agents

Reinforcement learning (RL) agents intended for use with gym-spectrum OpenAI Gym environments.

Overview

Todo

Dependencies

Written and tested on Python 3.6. Requires OpenAI Gym, Numpy, and Matplotlib.

Installation

It is recommend to create a venv.

python3 -m venv spectrum
source ./spectrum/bin/activate

To deactivate the venv, simply run the following in the shell:

deactivate

Direct from Git

You can clone the repo and install the module directly from your local Git clone using a pip editable installation.

git clone https://github.com/lukeprince20/spectrum-agents.git
cd spectrum-agents
pip install -e .

Supported Environments

The environments the spectrum agents are written to operate on are maintained in a separate repo: https://github.com/lukeprince20/gym-spectrum.

Agents

Currently agent support is varied per environment.

ChannelEnv Only

• Bandit
• Q-Learning

SpectrumEnv Only

• Genie
• Belief State Genie
• Replicated Q-Learning

Both

• No-op
• Incumbent
• Rotating
• Random