Safe adaptive importance sampling: a mixture approach

Central limit and other weak theorems (60F05) Martingales with continuous parameter (60G44) Sampling theory, sample surveys (62D05) Density estimation (62G07) Monte Carlo methods (65C05)

Abstract: This paper investigates adaptive importance sampling algorithms for which the policy, the sequence of distributions used to generate the particles, is a mixture distribution between a flexible kernel density estimate (based on the previous particles), and a "safe" heavy-tailed density. When the share of samples generated according to the safe density goes to zero but not too quickly, two results are established: (i) uniform convergence rates are derived for the policy toward the target density; (ii) a central limit theorem is obtained for the resulting integral estimates. The fact that the asymptotic variance is the same as the variance of an "oracle" procedure with variance-optimal policy, illustrates the benefits of the approach. In addition, a subsampling step (among the particles) can be conducted before constructing the kernel estimate in order to decrease the computational effort without altering the performance of the method. The practical behavior of the algorithms is illustrated in a simulation study.

Review text

Adaptive importance sampling (AIS) constitutes new samples, such as particles in statistical physics, generated under certain probability distribution called policy \(q_k\) and the next policy \(q_{k+1}\) uses the new particles adaptively. In the earlier works, the policy is chosen as the kernel density estimate based on the previous particles reweighted by their importance weights. The authors propose a new approach called `safe adaptive importance sampling' (SAIS) which estimates the policy as a mixture of kernel density estimate and certain `safe' density with heavier tails. They also consider the functional approximation and derive convergence rates, leading to a central limit theorem for the estimates. It is observed that the asymptotic variance with this procedure is the same as that of an `oracle' procedure. Further, a subsampling approach can be adopted to reduce the computational time involved without loosing the original efficiency. A simulation study at the end illustrates the practical nature of the algorithms developed. A section at the end gives detailed mathematical proofs including two appendices. There is a rich list of useful references.

Recommendations

Cites work

Cited in

(10)

This page was built for publication: Safe adaptive importance sampling: a mixture approach

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2039792)