3 ms·

A very interesting recent interview with an author of the concurrent work “decision transformer” https://open.spotify.com/episode/5uYwk4KcqRtdMjn7hlVOZD?si=2e4i

• algo_trader 4y ago
  Can you explain the main differences between DT and the submitted paper? DT is getting all the attention (ha! pun!)... maybe it is faster EDIT: ah. Here is a nice quote: Decision Transformer [5] bypasses the need for computing cumulative rewards through dynamic programming, but rather generates future actions conditioning on the desired returns, past states and actions