Thomas Lahore: This is a SUPER interesting idea!
Training Agents using Upside-Down Reinforcement Learning
"in upside-down RL ... the roles of actions and returns are switched."
5 replies, 308 likes
hardmaru: Training Agents using Upside-Down Reinforcement Learning
RL algorithms either predict rewards with value functions or maximize them using policy search. We study an alternative: Upside-Down RL, that solves RL problems primarily using supervised learning.
2 replies, 179 likes
NNAISENSE: Read them here:
Reinforcement Learning Upside Down: Don't Predict Rewards -- Just Map Them to Actions: https://arxiv.org/abs/1912.02875
Training Agents using Upside-Down Reinforcement Learning: https://arxiv.org/abs/1912.02877
0 replies, 23 likes
hardmaru: Companion paper by @rupspace et al., with some experiments. https://twitter.com/hardmaru/status/1203944037574594560?s=21
0 replies, 18 likes
HotComputerScience: Most popular computer science paper of the day:
"Training Agents using Upside-Down Reinforcement Learning"
0 replies, 1 likes
Found on Dec 09 2019 at https://arxiv.org/pdf/1912.02877.pdf