Exercise 6.4.2
Linear to deep function approximation: DQN, challenges, remedies
Reinforcement Learning › Exercises › Lecture 6 · Function Approximation › 6.4 Convergence, divergence and experience replay
Justify the transition from using a linear function approximator to a deep learning architecture, as seen in DQN. What new challenges arise from this increase in model complexity, and how are they managed algorithmically?