OpenAI researchers found that a simple generalization of classic count-based exploration to high-dimensional state spaces achieves near-SOTA performance in continuous control and Atari games. By estimating visitation frequencies via pseudo-counts, the method provides a simple yet powerful baseline for MDPs requiring extensive exploration.
Termsslug: mdp
MDP
马尔可夫决策过程
A mathematical framework for sequential decision-making problems, assuming future states depend only on the current state.
Curated Intelligence
OpenAI 官方博客T1·Models
68 pts
OpenAI 官方博客T1·Models
78 pts
OpenAI introduces RL², a framework that encodes a 'fast' reinforcement learning algorithm into the weights of a Recurrent Neural Network (RNN). By training this RNN with a general-purpose 'slow' RL algorithm across many tasks, it gains cross-episode memory, enabling rapid adaptation to new environments. Experiments demonstrate effectiveness on multi-armed bandits and large-scale visual navigation.