no code implementations • 11 Jun 2020 • Yuhua Zhu, Zach Izzo, Lexing Ying
The main idea is to borrow extra randomness from the future to approximately re-sample the next state when the underlying dynamics of the problem are sufficiently smooth.
