No description or website provided.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.circleci
data/maze0/expert_human
docs
irl_benchmark
.gitignore
CHANGES.txt
DEVELOPER.md
LICENSE
MANIFEST.in
README.md
__init__.py
generate_expert_data.py
main.py
pylintrc
requirements.txt
setup.py

README.md

CircleCI Maintainability Test Coverage

Introduction

irl-benchmark is a modular library for evaluating various Inverse Reinforcement Learning algorithms. It provides an extensible platform for experimenting with different environments, algorithms and metrics.

Installation

conda create --name irl-benchmark python=3.6

source activate irl-benchmark

pip install -r requirements.txt

Getting Started

Start by generating expert data by

python generate_expert_data.py

Then run

python main.py

to get an overview of how all the components of irl-benchmark work together.

Documentation

Documentation is available as work in progress at: https://johannesheidecke.github.io/irl-benchmark.

You may find the extending part useful if you are planning to author new algorithms.

Environemts

Algorithms

Metrics

Copyright: Adria Garriga-Alonso, Anton Osika, Johannes Heidecke, Max Daniel, and Sayan Sarkar.