Bellman backup

New analytic planning method tackles uncertainty in reinforcement learning

Researchers have developed a new method for model-based reinforcement learning in stochastic environments that accounts for predictive uncertainty. This approach uses a quadratic action-value parameterization to simplif…