Skip to content
ArcadeLearningEnvironment Julia interface
Branch: master
Clone or download
Pull request Compare This branch is 20 commits ahead of nowozin:master.
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.
deps
src
test
.gitignore
.travis.yml
LICENSE.md
README.md
REQUIRE
appveyor.yml

README.md

Build Status codecov

ArcadeLearningEnvironment.jl

This package is a Julia wrapper for the ArcadeLearningEnvironment (ALE).

ALE is a modified emulator for the Atari 2600 that can emulate around 50 games with additional access to game state information and in-game rewards. This is useful for learning and benchmarking artificial intelligence agents playing computer games.

Citation

If you use this package for research publications, please cite the following paper to acknowledge the work that went into ALE.

@ARTICLE{bellemare13arcade,
	author = {{Bellemare}, M.~G. and {Naddaf}, Y. and {Veness}, J. and {Bowling}, M.},
	title = {The Arcade Learning Environment: An Evaluation Platform for General Agents},
	journal = {Journal of Artificial Intelligence Research},
	year = 2013,
	month = 06,
	volume = 47,
	pages = {253--279}
}

Installation

On Mac OS and Linux the package automatically downloads and builds version 0.6.0 of the ArcadeLearningEnvironment by adding it in julia 0.6 with

Pkg.add("ArcadeLearningEnvironment")

or in the package REPL of julia 0.7.0 with

add ArcadeLearningEnvironment

On Windows (which I have not tried yet) you can build the libale_c.dll file manually and set the LIBALE_HOME environment variable to the directory containing this file. Then, the above two commands should work as well. Note that this is untested and any correction or feedback is welcome.

Example

using ArcadeLearningEnvironment


episodes = 50

ale = ALE_new()
loadROM(ale, "seaquest")

S = zeros(Int64, episodes)
TR = zeros(episodes)
for ei = 1:episodes
    ctr = 0.0

    fc = 0
    while game_over(ale) == false
        actions = getLegalActionSet(ale)
        ctr += act(ale, actions[rand(1:length(actions))])
        fc += 1
    end
    reset_game(ale)
    println("Game $ei ended after $fc frames with total reward $(ctr).")

    S[ei] = fc
    TR[ei] = ctr
end
ALE_del(ale)
You can’t perform that action at this time.