Trajectory transformer

tags

Transformers, Reinforcement learning, GPT

paper

(Janner et al. 2021)

Architecture

It is a similar model to Decision transformer, with some added techniques to encode a trajectory.

Bibliography

1. Michael Janner, Qiyang Li, Sergey Levine. November 28, 2021. "Offline Reinforcement Learning as One Big Sequence Modeling Problem". arXiv. DOI.