Gradient-free Hamiltonian Monte Carlo with Efficient Kernel Exponential Families
This addresses a bottleneck for practitioners in Bayesian inference and MCMC when gradients are unavailable, offering an incremental advancement over existing gradient-free methods.
The paper tackles the problem of sampling from target densities with intractable gradients by proposing Kernel Hamiltonian Monte Carlo (KMC), a gradient-free adaptive MCMC algorithm that learns gradient structure using kernel exponential families, resulting in substantial mixing improvements over state-of-the-art gradient-free samplers.
We propose Kernel Hamiltonian Monte Carlo (KMC), a gradient-free adaptive MCMC algorithm based on Hamiltonian Monte Carlo (HMC). On target densities where classical HMC is not an option due to intractable gradients, KMC adaptively learns the target's gradient structure by fitting an exponential family model in a Reproducing Kernel Hilbert Space. Computational costs are reduced by two novel efficient approximations to this gradient. While being asymptotically exact, KMC mimics HMC in terms of sampling efficiency, and offers substantial mixing improvements over state-of-the-art gradient free samplers. We support our claims with experimental studies on both toy and real-world applications, including Approximate Bayesian Computation and exact-approximate MCMC.