Correlated Equilibria in Extensive Form Games

Features

• Specify multi-player extensive form games as tree structures, with potentially imperfect information
  • Convert from extensive form to normal (strategic) form games
  • Convert from extensive form to sequence form games
• Compute correlated equilibria using linear programming techniques
  • Simplex
  • Lemke-Howlson
• Reinforcement Learning Algorithms
  • Weighted Majority
  • Perturbed Follow-the-Leader
  • Regularized Follow-the-Leader

Coming Soon in Feb. 2020

• Compute correlated equilibria using counterfactual regret minimization algorithms

Reference Papers

• Correlated Q-Learning (Greenwald and Hall, 2003)
• Combining No-regret and Q-learning (Kash, Sullins, and Hofmann, 2019)
• No-Regret Learning in Convex Games (Gordon, Greenwald, and Marks, 2008)