Lilian Weng • 6/23/2019

Meta Reinforcement Learning

This technical article delves into meta reinforcement learning (Meta-RL), explaining how agents can be trained to generalize and adapt rapidly to new reinforcement learning tasks. It traces the concept's origins from a 2001 paper to modern implementations like RL^2, discussing the training over distributions of MDPs and the goal of creating general-purpose learning algorithms.

0 comments

#Machine Learning #artificial intelligence #Meta Learning