Papers of the day   All papers

“Superstition” in the Network: Deep Reinforcement Learning Plays Deceptive Games

Comments

Aug 15 2019 Julian Togelius

Can deep networks be superstitious? We trained deep networks with the standard A2C deep reinforcement learning algorithm to play "deceptive" games and found a range of behaviors that could only be described as bizarre. and, yes, superstitious. https://arxiv.org/abs/1908.04436 https://t.co/Obxy4AMVrV
6 replies, 287 likes


Aug 15 2019 hardmaru

Superstition in the Network: Deep Reinforcement Learning Plays Deceptive Games 👻 https://arxiv.org/abs/1908.04436
1 replies, 103 likes


Oct 11 2019 Julian Togelius

Paper here: https://arxiv.org/abs/1908.04436
1 replies, 27 likes


Aug 15 2019 Julian Togelius

Our paper, "Superstition in the Network: Deep Reinforcement Learning Plays Deceptive Games" is co-authored by @FilipoGiovanni @Amidos2006 @DamorinSolusar @matthew_stephe @ChristophSalge and me, and will be presented at @AIIDEconference in October. https://arxiv.org/abs/1908.04436
1 replies, 25 likes


Sep 01 2019 Janislav Jankov

Brief summary of "Superstition" in the Network by @FilipoGiovanni, @Amidos2006, D Anderson, M Stephenson, @ChristophSalge, @togelius https://arxiv.org/abs/1908.04436 Researchers torment RL agent on deceptive games. The agent learns to do a make-it-rain "ritual" on one of them.1/10
1 replies, 17 likes


Aug 15 2019 Philip Bontrager

Cause and Effect is hard. Here we train an agent to invest with bankers and then wait for it’s riches. Our agent likes to stand in its lucky corner, where it happens to be, when the reward comes in. I think we’ve almost cracked human intelligence.
1 replies, 16 likes


Sep 01 2019 Julian Togelius

Nice summary of our recent "'Superstition' in the network" paper.
0 replies, 14 likes


Aug 15 2019 IntuitionMachine

Deep Learning research never fails to amaze. Now you can train networks to be superstitious!!! https://twitter.com/togelius/status/1161975295810949120
1 replies, 5 likes


Oct 11 2019 AIIDE Conference

We're listening to the talk on “Superstition’” in the Network: Deep Reinforcement Learning Plays Deceptive Games discussing the various "deceptions" reinforcement learning algorithms fall for, proposing a typology of deceptions and illuminating the character of the algorithms
0 replies, 4 likes


Aug 14 2019 Brundage Bot

Superstition in the Network: Deep Reinforcement Learning Plays Deceptive Games. Philip Bontrager, Ahmed Khalifa, Damien Anderson, Matthew Stephenson, Christoph Salge, and Julian Togelius http://arxiv.org/abs/1908.04436
1 replies, 1 likes


Feb 11 2020 HotComputerScience

Most popular computer science paper of the day: "“Superstition” in the Network: Deep Reinforcement Learning Plays Deceptive Games" https://hotcomputerscience.com/paper/-superstition-in-the-network-deep-reinforcement-learning-plays-deceptive-games https://twitter.com/togelius/status/1161975295810949120
0 replies, 1 likes


Content