A generally intelligent agent must be able to teach itself how to solve
problems in complex domains with minimal human supervision. Recently, deep
reinforcement learning algorithms combined with self-play have achieved
superhuman proficiency in Go, Chess, and Shogi without human data or domain
knowledge. In these environments, a reward is always received at the end of the
game, however, for many combinatorial optimization environments, rewards are
sparse and episodes are not guaranteed to terminate. We introduce Autodidactic
Iteration: a novel reinforcement learning algorithm that is able to teach
itself how to solve the Rubik's Cube with no human assistance. Our algorithm is
able to solve 100% of randomly scrambled cubes while achieving a median solve
length of 30 moves -- less than or equal to solvers that employ human domain
knowledge.
Description
[1805.07470] Solving the Rubik's Cube Without Human Knowledge
%0 Generic
%1 mcaleer2018solving
%A McAleer, Stephen
%A Agostinelli, Forest
%A Shmakov, Alexander
%A Baldi, Pierre
%D 2018
%K 2018 arxiv games paper reinforcement-learning
%T Solving the Rubik's Cube Without Human Knowledge
%U http://arxiv.org/abs/1805.07470
%X A generally intelligent agent must be able to teach itself how to solve
problems in complex domains with minimal human supervision. Recently, deep
reinforcement learning algorithms combined with self-play have achieved
superhuman proficiency in Go, Chess, and Shogi without human data or domain
knowledge. In these environments, a reward is always received at the end of the
game, however, for many combinatorial optimization environments, rewards are
sparse and episodes are not guaranteed to terminate. We introduce Autodidactic
Iteration: a novel reinforcement learning algorithm that is able to teach
itself how to solve the Rubik's Cube with no human assistance. Our algorithm is
able to solve 100% of randomly scrambled cubes while achieving a median solve
length of 30 moves -- less than or equal to solvers that employ human domain
knowledge.
@misc{mcaleer2018solving,
abstract = {A generally intelligent agent must be able to teach itself how to solve
problems in complex domains with minimal human supervision. Recently, deep
reinforcement learning algorithms combined with self-play have achieved
superhuman proficiency in Go, Chess, and Shogi without human data or domain
knowledge. In these environments, a reward is always received at the end of the
game, however, for many combinatorial optimization environments, rewards are
sparse and episodes are not guaranteed to terminate. We introduce Autodidactic
Iteration: a novel reinforcement learning algorithm that is able to teach
itself how to solve the Rubik's Cube with no human assistance. Our algorithm is
able to solve 100% of randomly scrambled cubes while achieving a median solve
length of 30 moves -- less than or equal to solvers that employ human domain
knowledge.},
added-at = {2018-06-17T20:23:21.000+0200},
author = {McAleer, Stephen and Agostinelli, Forest and Shmakov, Alexander and Baldi, Pierre},
biburl = {https://www.bibsonomy.org/bibtex/246608c1bc89217deb14c3340ce0447bf/achakraborty},
description = {[1805.07470] Solving the Rubik's Cube Without Human Knowledge},
interhash = {965666151bbd287f4d059b3127ece38a},
intrahash = {46608c1bc89217deb14c3340ce0447bf},
keywords = {2018 arxiv games paper reinforcement-learning},
note = {cite arxiv:1805.07470Comment: First three authors contributed equally. Submitted to NIPS 2018},
timestamp = {2018-06-17T20:23:21.000+0200},
title = {Solving the Rubik's Cube Without Human Knowledge},
url = {http://arxiv.org/abs/1805.07470},
year = 2018
}