MDP Planning as Policy Inference

Name: MDP Planning as Policy Inference
Uploaded: 2026-08-20T12:45:27.002394+00:00
Duration: 744 s
Description: The episode discusses "MDP Planning as Policy Inference," which reframes classic planning as Bayesian inference over policies. Instead of finding a single optimal action, the method derives a distribution over possible strategies, providing a measure of uncertainty. The technique uses variational sequential Monte Carlo (VSMC) and shows differences from Soft Actor-Critic (SAC) baselines.

The episode discusses "MDP Planning as Policy Inference," which reframes classic planning as Bayesian inference over policies. Instead of finding a single optimal action, the method derives a distribution over possible strategies, providing a measure of uncertainty. The technique uses variational sequential Monte Carlo (VSMC) and shows differences from Soft Actor-Critic (SAC) baselines.

Open the full episode