multiplicative-weights

Rock paper scissors against a Hedge ensemble that learns your habits — beats patterned players 80-100% while correctly settling at chance against true randomness.

CPU-first experiments on predictive correspondences between adaptive dynamical systems

This rep contains the projects made for the course "Reinforcement Learning and Dynamic Optimization" at TUC (2024).

Solver for repeated payoff-matrix games that adapts to non-stationary opponents.