7406f6f11e
- Maximum Entropy 모델 - Bayesian Brain 이론 - Predictive Coding - Free Energy Principle - MemGPT 구현 - ANN Search 알고리즘 - Information Theory와 메모리 - Trust & Reputation 시스템 - Actor-Critic 강화학습 - 최신 메모리 시스템 서베이
617 B
617 B
Actor-Critic Algorithms
- Authors: R. S. Sutton, A. G. Barto
- Year: 1998 (in Reinforcement Learning: An Introduction)
- Summary: This chapter details Actor-Critic methods in reinforcement learning. The 'Actor' learns a policy (what to do), while the 'Critic' learns a value function (how good the action was). This can be applied to memory retrieval, where the Actor decides which memory to recall and the Critic evaluates how useful that recall was for the current task, allowing the agent to learn better retrieval strategies over time.
- Link: http://incompleteideas.net/book/the-book-2nd.html