Google Deepmind DQN plays Atari Pacman

Показать описание

Google Deepminds deep learning algorithm DQN learns to play Atari Pacman. You can clearly see progress, although learning time was not that long. The algorithm knew nothing about the controls. It was fed with raw pixel data and had to decide on the next action. It had no idea of the princible of a wall or a ghost. It's only goal was to maximize the score.

Рекомендации по теме

Комментарии

If deep mind can't even beat Pacman after a day of playing I feel better about myself.

CoffeeTrav

The problem with this game is
That the a.i can't compare 2 different states,
It does not know that it just made progress by collencting one of these white dots, it only sees which movements are ideal in most cases in a similar situation.

You would need an a.i that is rewarded while playing not only at the end for a highscore.

Also finding a connection between the screen having more black/ that correlating with a higher score/ and how it achieves the screens blackness seem incomprehensible for this type of A.i

dmarsub

Why Pacman doesn't eat the ghosts?

domenicolattuca

Wouldn't it make sense to punish the fitness when he dies?

Speedow

That's very cool! I want to ask what is the experiment platform, whether is open source. Thanks!

yongliu

Google Deepmind DQN plays Atari Pacman

Google DeepMind's Deep Q-learning playing Atari Breakout!

Google Deepmind DQN plays Atari Breakout

Globalfuturist.org: Google #DeepMind Deep Q learning playing Atari Breakout

Google Deepmind DQN plays Atari Pacman

Google DeepMind's Deep Q learning playing Atari Breakout

Google DeepMind Artari Plays Pacman

Google DeepMind's Deep Q-Learning & Superhuman Atari Gameplays | Two Minute Papers #27

Simple analysis: Google DeepMind Artari Plays Pacman

Deep Reinforcement learning | Deep Q learning for atari games Deepmind |

Deep Reinforcement Learning: Agent Playing Atari

Google DeepMind's Deep Q-learning playing Atari Breakout

[Classic] Playing Atari with Deep Reinforcement Learning (Paper Explained)

DeepMind DQN Tutorial - Installation and explanation of the Atari-playing agent

Google DeepMind Artari Plays Atlantis

Google's DeepMind AI taught itself to Walk, Run & Jump

Deep Reinforcement Learning Agent Playing Atari

AI learns to play ATARI and DOOM - deep reinforcement learning experiments

Deep Reinforcement Learning in a 3-D Blockworld Environment

Deep Q-Learning - Atari Breakout

google DQN deep q network on atari 2600 pong game

Playing Atari with Deep Reinforcement Learning - TWiML Online Meetup - March 2018

The Future Starts Here: Interview with DeepMind – Google's AI research lab

Agent57: Outperforming the Atari Human Benchmark

Deepmind artificial intelligence @ FDOT14