Generalization and Regularization in DQN

Framing and objectives Problem statement: brittleness in value‑based agents At first glance the work addresses a glaring mismatch: contemporary Deep Q-Networks (DQN) can achieve high scores yet struggle outside training conditions, suggesting limited...