Marrying Hamiltonian Monte Carlo with annealed importance sampling results in an unbiased estimate of marginal likelihood that supports pathwise derivatives. Surprisingly, stochastic gradient annealed importance sampling turns out to be inconsistent.
Guodong Zhang,
Kyle Hsu,
Jianing Li,
Chelsea Finn,
Roger Grosse
Efficient controller synthesis for reachability and safety specifications via on-demand abstraction construction and adaptive spatiotemporal resolution.