Chapter
Long-term Planning in Bayesian Frameworks for Optimal Decision Making
58:16 - 1:03:42 (05:26)
The Bayesian framework with prior of possible worlds and long-term planning can provide optimal decision making with the right amount of exploration, important for simple problems such as the bandit problem.
Clips
The Markov assumption is common in reinforcement learning as it makes the mathematics much easier to deal with.
58:16 - 1:01:07 (02:51)
Summary
The Markov assumption is common in reinforcement learning as it makes the mathematics much easier to deal with. This assumption states that the next observation only depends on the previous observation, simplifying the learning process.
ChapterLong-term Planning in Bayesian Frameworks for Optimal Decision Making
Episode#75 – Marcus Hutter: Universal Artificial Intelligence, AIXI, and AGI
PodcastLex Fridman Podcast
The Bayesian framework, which takes a prior of possible worlds and combines it with Bayes optimal decision-making, automatically implies the right amount of exploration, not too much and not too little.
1:01:07 - 1:03:42 (02:34)
Summary
The Bayesian framework, which takes a prior of possible worlds and combines it with Bayes optimal decision-making, automatically implies the right amount of exploration, not too much and not too little. Exploration is important for learning and gaining new knowledge, while planning is important for long-term decision making.