Hyperbolic Inverse Reinforcement Learning
Hyperbolic distance parameterization of rewards boosts IRL accuracy in low-data regimes and reveals non-separable conjunctive structures.
Stories per day · 30 days
1 stories, newest first · one card per story