Deep Neural Networks Tend To Extrapolate Predictably.

Katie Kang, Amrith Setlur, Claire J. Tomlin, Sergey Levine

Browse the full ICLR paper archive.