Papers of the day   All papers

Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions

Comments

Jeff Clune: Introducing Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions.Adds domain-general metrics of environment novelty and progress in open-ended algorithms + a new environment search space https://arxiv.org/abs/2003.08536 1/ https://t.co/jKSfQlHDXP

4 replies, 229 likes


Rui Wang: Thrilled to release a major enhancement of the POET algorithm - "Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges and their Solutions." paper: http://arxiv.org/abs/2003.08536 w/ @joelbot3000 @AdityaRawaI @_calio Yulun Li @jeffclune and @kenneth0stanley https://t.co/gQjNaKqXR2

4 replies, 114 likes


hardmaru: You vs the hardcore bipedal robot she told you not to worry about👇🏼

2 replies, 111 likes


Kenneth Stanley: Excited to announce a major enhancement of the POET algorithm - broadening and simplifying its potential application to make almost any conceivable domain open-ended. With @ruiwang2uiuc @joelbot3000 @jeffclune @AdityaRawaI @_calio Yulun Li and @jeffclune https://arxiv.org/abs/2003.08536 https://t.co/yTt2a4vutM

1 replies, 99 likes


Rui Wang: Excited to announce: source code for Enhanced POET now available at http://github.com/uber-research/poet paper: http://arxiv.org/abs/2003.08536 w/ @joelbot3000 @AdityaRawaI @_calio Yulun Li @jeffclune and @kenneth0stanley Distributed computing in our code enabled by Fiber: http://uber.github.io/fiber/ https://t.co/XaHQGQaoLf

0 replies, 96 likes


Kenneth Stanley: Happy to share that our paper "Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges & their Solutions" will appear at #icml2020 https://arxiv.org/abs/2003.08536 Led by @ruiwang2uiuc w/ @joelbot3000, @AdityaRawaI, @_calio, Yulun Li, @jeffclune, and myself 1/2 https://t.co/j7Iemf2s4a

1 replies, 66 likes


Natasha Jaques: How novel is a new environment/task? This paper suggests a new metric based on whether it introduces a significantly new ranking over agents compared to previous environments. Pretty interesting!

0 replies, 51 likes


Jeff Clune: A blog post is now out for Enhanced POET: https://eng.uber.com/enhanced-poet-machine-learning, which adds domain-general metrics of environment novelty and progress in open-ended algorithms + a new environment search space https://arxiv.org/abs/2003.08536 Progress on AI-GAs Pillar 3 (auto-generating environments) 1/ https://t.co/RavYtMlKO8

2 replies, 41 likes


Joel Lehman: Excited to be involved w/ “Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges and their Solutions” (check out video) -- great work led by @ruiwang2uiuc w/ @AdityaRawaI @_calio Yulun Li @jeffclune and @kenneth0stanley https://arxiv.org/abs/2003.08536 https://t.co/QQ4BhjUB7V

0 replies, 39 likes


Aditya Rawal: New blog post on Enhanced POET is up at: https://eng.uber.com/enhanced-poet-machine-learning, Arxiv: https://arxiv.org/abs/2003.08536 Work led by @ruiwang2uiuc w/ @joelbot3000, myself, @_calio, Yulun Li, @jeffclune, and @kenneth0stanley. https://t.co/xATGqihvzA

0 replies, 30 likes


Rui Wang: Look forward to presenting our most recent work on Open-endedness at #icml2020! "Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges & their Solutions" http://arxiv.org/abs/2003.08536 w/ @joelbot3000 @AdityaRawaI @_calio Yulun @jeffclune @kenneth0stanley 1/2 https://t.co/hlpo6Nw4dB

1 replies, 25 likes


Joel Lehman: New blog post on Enhanced POET is up at: https://eng.uber.com/enhanced-poet-machine-learning -- open-ended search for reinforcement learning through inventing and solving problems. Arxiv: https://arxiv.org/abs/2003.08536, work led by @ruiwang2uiuc w/ me @AdityaRawaI @_calio Yulun Li @jeffclune @kenneth0stanley

0 replies, 23 likes


Rui Wang: Have not got time to read the full Enhanced POET paper? We’ve just released a @UberEng blog post about it at http://eng.uber.com/enhanced-poet-machine-learning ! paper: http://arxiv.org/abs/2003.08536 code: http://github.com/uber-research/poet w/ @joelbot3000 @AdityaRawaI @_calio Yulun Li @jeffclune and @kenneth0stanley https://t.co/5P6zxb82lF

0 replies, 22 likes


Aditya Rawal: Introducing our new work “Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges and their Solutions” -- Work led by @ruiwang2uiuc w/ @joelbot3000, @_calio, Yulun Li, @jeffclune, @kenneth0stanley, and myself. https://arxiv.org/abs/2003.08536 https://t.co/YnOCrxewSt

0 replies, 21 likes


Aditya Rawal: Excited to share that our paper "Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges & their Solutions" will appear at #icml2020 https://arxiv.org/abs/2003.08536 Led by @ruiwang2uiuc w/ @joelbot3000, @_calio, Yulun Li, @jeffclune and @kenneth0stanley

0 replies, 20 likes


Eugene Vinitsky: Come for the music, stay for a wild set of generated environments. Idea here is to co-evolve agents and environments to construct an increasingly complex curriculum New tricks: - Encode terrain height as output of neural net. Nice parametrization.

1 replies, 19 likes


Jeff Clune: Population-based methods are rapidly being adopted by the ML community (tutorial: https://www.youtube.com/watch?v=g6HiuEnbwJE). Fiber enabled our recent Enhanced POET https://twitter.com/jeffclune/status/1241016690680270849 and our new paper "Scaling MAP-Elites to Deep Neuroevolution" https://arxiv.org/abs/2003.01825

0 replies, 16 likes


Jiale Zhi: Excited to announce the release of Enhanced POET code! The computation is fully powered by Fiber, which scaled the algorithm to 10+x computational resources than ipyparallel! https://uber.github.io/fiber/ http://github.com/uber-research/poet

1 replies, 15 likes


Jiale Zhi: Excited to share our new blog post on Enhanced POET: https://eng.uber.com/enhanced-poet-machine-learning in case you didn't have time to read the paper https://arxiv.org/abs/2003.08536 Work led by @ruiwang2uiuc w/ @joelbot3000, @AdityaRawaI, myself, Yulun Li, @jeffclune and @kenneth0stanley. https://t.co/kjfvc5mCaX

0 replies, 13 likes


hardmaru: @ayush1292 @erwincoumans Would be a cool Gym task if the terrain is randomly generated according to some difficulty level. I think @ruiwang2uiuc might be interested! https://twitter.com/ruiwang2uiuc/status/1251164246999588883 https://t.co/BYurW59ysg

1 replies, 8 likes


Jiale Zhi: Check out our new #icml2020 paper "Enhanced POET: Open-ended RL through Unbounded Invention of Learning Challenges & their Solutions" http://arxiv.org/abs/2003.08536 https://eng.uber.com/enhanced-poet-machine-learning Led by @ruiwang2uiuc w/ @joelbot3000 @AdityaRawaI myself Yulun Li @jeffclune and @kenneth0stanley https://t.co/5UUjJc7nOh

0 replies, 7 likes


Berick: If all the state of the art AI algorithms jump off a cliff, will yours? #BadParentingGoodAI

1 replies, 4 likes


Yannic Kilcher: Code is out for the enhanced POET. My video about it: https://youtu.be/gbG1X8Xq-T8 ML street talk: https://youtu.be/sIv-JoFf-Ww

0 replies, 3 likes


MONTREAL.AI: Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions Wang et al.: https://arxiv.org/abs/2003.08536 #OpenEnded #ReinforcementLearning https://t.co/09hZyBnAUG

0 replies, 2 likes


Calio: Excited to share our new work "Enhanced POET", an algorithm that learns through unbounded invention of challenges and solutions. Great work led by @ruiwang2uiuc with @joelbot3000 @jeffclune @AdityaRawaI Yulun Li @jeffclune and @kenneth0stanley https://arxiv.org/abs/2003.08536 https://t.co/XhuzmbrWw7

0 replies, 1 likes


HotComputerScience: Most popular computer science paper of the day: "Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions" https://hotcomputerscience.com/paper/enhanced-poet-open-ended-reinforcement-learning-through-unbounded-invention-of-learning-challenges-and-their-solutions https://twitter.com/jeffclune/status/1241016690680270849

0 replies, 1 likes


Brundage Bot: Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions. Rui Wang, Joel Lehman, Aditya Rawal, Jiale Zhi, Yulun Li, Jeff Clune, and Kenneth O. Stanley http://arxiv.org/abs/2003.08536

1 replies, 1 likes


Content

Found on Mar 20 2020 at https://arxiv.org/pdf/2003.08536.pdf

PDF content of a computer science paper: Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions