%0 Generic %1 srivastava2019training %A Srivastava, Rupesh Kumar %A Shyam, Pranav %A Mutz, Filipe %A Jaśkowski, Wojciech %A Schmidhuber, Jürgen %D 2019 %K 2019 arxiv reinforcement-learning %T Training Agents using Upside-Down Reinforcement Learning %U http://arxiv.org/abs/1912.02877 %X Traditional Reinforcement Learning (RL) algorithms either predict rewards with value functions or maximize them using policy search. We study an alternative: Upside-Down Reinforcement Learning (Upside-Down RL or UDRL), that solves RL problems primarily using supervised learning techniques. Many of its main principles are outlined in a companion report 34. Here we present the first concrete implementation of UDRL and demonstrate its feasibility on certain episodic learning problems. Experimental results show that its performance can be surprisingly competitive with, and even exceed that of traditional baseline algorithms developed over decades of research.

@misc{srivastava2019training, abstract = {Traditional Reinforcement Learning (RL) algorithms either predict rewards with value functions or maximize them using policy search. We study an alternative: Upside-Down Reinforcement Learning (Upside-Down RL or UDRL), that solves RL problems primarily using supervised learning techniques. Many of its main principles are outlined in a companion report [34]. Here we present the first concrete implementation of UDRL and demonstrate its feasibility on certain episodic learning problems. Experimental results show that its performance can be surprisingly competitive with, and even exceed that of traditional baseline algorithms developed over decades of research.}, added-at = {2019-12-10T08:56:22.000+0100}, author = {Srivastava, Rupesh Kumar and Shyam, Pranav and Mutz, Filipe and Jaśkowski, Wojciech and Schmidhuber, Jürgen}, biburl = {https://www.bibsonomy.org/bibtex/23a996d1858046143046e52aa9de3491b/analyst}, description = {[1912.02877] Training Agents using Upside-Down Reinforcement Learning}, interhash = {27ae7733112ad6c9d060a4d6c92fe895}, intrahash = {3a996d1858046143046e52aa9de3491b}, keywords = {2019 arxiv reinforcement-learning}, note = {cite arxiv:1912.02877Comment: NNAISENSE Technical Report. 17 pages, 6 figures}, timestamp = {2019-12-10T08:56:22.000+0100}, title = {Training Agents using Upside-Down Reinforcement Learning}, url = {http://arxiv.org/abs/1912.02877}, year = 2019 }