Studies evolution and learning. The discussion includes fictitious play and replicator dynamics, and stochastic stability is considered at length. Following this, some detailed discussion of stochastic stability is given, including the computation of invariant distributions and minimum cost trees. Blackwell approachability is used to define strategies that minimize regret across all actions. This is then connected to correlated equilibrium. Calibrated forecasts are defined, and a connection to correlated equilibria also noted. Provides a brief discussion of Bayesian learning and the key role...

Studies evolution and learning. The discussion includes fictitious play and replicator dynamics, and stochastic stability is considered at length. Following this, some detailed discussion of stochastic stability is given, including the computation of invariant distributions and minimum cost trees. Blackwell approachability is used to define strategies that minimize regret across all actions. This is then connected to correlated equilibrium. Calibrated forecasts are defined, and a connection to correlated equilibria also noted. Provides a brief discussion of Bayesian learning and the key role of the martingale convergence theorem.