Oracle-based Uniform Sampling from Convex Bodies
By: Thanh Dang, Jiaming Liang
Potential Business Impact:
Helps computers find hidden patterns in data faster.
We propose new Markov chain Monte Carlo algorithms to sample a uniform distribution on a convex body $K$. Our algorithms are based on the Alternating Sampling Framework/proximal sampler, which uses Gibbs sampling on an augmented distribution and assumes access to the so-called restricted Gaussian oracle (RGO). The key contribution of this work is the efficient implementation of RGO for uniform sampling on $K$ via rejection sampling and access to either a projection oracle or a separation oracle on $K$. In both oracle cases, we establish non-asymptotic complexities to obtain unbiased samples where the accuracy is measured in R\'enyi divergence or $\chi^2$-divergence.
Similar Papers
Convergence of a Sequential Monte Carlo algorithm towards multimodal distributions on Rd
Computation
Helps computers find patterns in complex data.
Convergence of a Sequential Monte Carlo algorithm towards multimodal distributions on Rd
Computation
Helps computers find patterns in complex data.
When Langevin Monte Carlo Meets Randomization: Non-asymptotic Error Bounds beyond Log-Concavity and Gradient Lipschitzness
Machine Learning (Stat)
Makes computer models work better for hard problems.