Welcome to deepatari!

This is a framework to support the development of algorithms for intelligent agents in the Atari game playing domain. The idea is to let researcher or programmer focus on the development of algorithms without having to think about the whole environment implementation and to facilitate and generalize the evaluation of these algorithms to make results comparable.

Help

• Installation guide
  • Create a working directory
  • Prepare conda environment
  • Add non standard packages
  • Install deepatari

API reference

• Experiment
  • Experiment - Base class
  • AtariExp
• Agent
  • Agent - Base class
  • AtariAgent
• Environment
  • Environment - Base class
  • AtariEnv
• Learner
  • Learner - Base class
  • DQNNeon
  • DQNLasagne
• Memory
  • Memory - Base class
  • ReplayMemory
• Statistics

About

• License
• Authors
• Changelog
  • Version 0.1
  • Version 0.0

Indices and tables

• Index
• Module Index
• Search Page