Skip to content

sachin-101/OpenAI-Requests-for-Research

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 

Repository files navigation

OpenAI Requests For Research 2.0

OpenAI has released a new batch of seven unsolved problems which have come up in the course of their research at OpenAI. These problems are a fun and meaningful way for new people to enter the field, as well as for practitioners to hone their skills. Many will require inventing new ideas.

Warmups

  • Train an LSTM to solve xor problem.
  • Implement a clone of classic Sanke Game and solve it using RL algorithm.

Main Problem statements

  • Slitherin'. : Implement and solve a multiplayer clone of classic Snake game. (Work Going On)
  • Parameter Averaging in Distributed RL
  • Transfer learning between different games via Generative Models
  • Transformers with Linear Attention
  • Learned Data Augmentation
  • Regularization in Reinforcement Learning
  • Automated Solutions of Olympiad Inequality Problems

Training Agent

Training Agent

  • Using Double DQN(RL algo) to train the AI agent.
  • Using Curriculum Learning approach to train the agent. Basically, keeping the playing area small at the beginning of training. And increasing the size of playing area as the training progresses.
  • This helps to solve the problem of Sparse Reward, cause if the agent starts in a very big playing area, it becomes very difficult for it to find positive reward in th environment, thus disabling the agent from learning the optimal solution.
  • Small playing area ensures that the agent bumps into the food often, even while performing random actions(epsilon close to 1).

Learned Agent

Learned Agent

Releases

No releases published

Packages

 
 
 

Contributors

Languages