Advanced Bandit Algorithms Research

Computational models of cognitive stages show that network-stage architectures outperform linear stages through estimation efficiency and explicit exposure tracking, not just parameter reduction.

Contextual Thompson sampling approach for personalizing exercise sequences in digital learning environments, optimizing skill advancement at scale using bandit-based algorithms.

Study reveals NFL sportsbooks misprice money line underdog odds relative to spread odds, enabling conditional returns up to 6.55% and challenging market efficiency assumptions.