Non-exponential Reward Discounting in Reinforcement Learning

Reinforcement learning methods typically discount future rewards using an exponential scheme to achieve theoretical convergence guarantees. Studies from neuroscience, psychology, and economics suggest that human animal behavior is better captured by the hyperbolic discounting model. Hyperbolic has recently been studied in deep reinforcement shown promising results. However, this area of research seemingly understudied, with most extant continuing standard formulation. My dissertation examines effects non-exponential functions (such as hyperbolic) on agent's aims investigate their impact multi-agent systems generalization tasks. A key objective study link rate approximation underlying hazard its environment through survival analysis.