Abstract: The empirical success of derivative-free methods in reinforcement learning for planning through contact seems at odds with the perceived fragility of classical gradient-based optimization ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback