AND/OR Importance Sampling
نویسندگان
چکیده
The paper introduces AND/OR importance sampling for probabilistic graphical models. In contrast to importance sampling, AND/OR importance sampling caches samples in the AND/OR space and then extracts a new sample mean from the stored samples. We prove that AND/OR importance sampling may have lower variance than importance sampling; thereby providing a theoretical justification for preferring it over importance sampling. Our empirical evaluation demonstrates that AND/OR importance sampling is far more accurate than importance sampling in many cases.
منابع مشابه
Learning When to Reject an Importance Sample
When observations are incomplete or data are missing, approximate inference methods based on importance sampling are often used. Unfortunately, when the target and proposal distributions are dissimilar, the sampling procedure leads to biased estimates or requires a prohibitive number of samples. Our method approximates a multivariate target distribution by sampling from an existing, sequential ...
متن کاملStratified Sampling with Spherically Symmetric Importance Samplers
SUMMARY Multivariate normal or Student importance sampling is a commonly used technique for integration problems in statistical inference. This integration approach is easy to implement, has straightforward error estimates and is eeective in a number of problems. A variety of variance reduction techniques can be considered with importance sampling. Stratiied sampling is one of these and in fact...
متن کاملAn overview of importance splitting for rare event simulation
Monte Carlo simulations are a classical tool to analyse physical systems. When unlikely events are to be simulated, the importance sampling technique is often used instead of Monte Carlo. Importance sampling has some drawbacks when the problem dimensionality is high or when the optimal importance sampling density is complex to obtain. In this paper, we focus on a quite novel but somehow confide...
متن کاملImproving Importance Sampling by Adaptive Split-Rejection Control in Bayesian Networks
Importance sampling-based algorithms are a popular alternative when Bayesian network models are too large or too complex for exact algorithms. However, importance sampling is sensitive to the quality of the importance function. A bad importance function often leads to much oscillation in the sample weights, and, hence, poor estimation of the posterior probability distribution. To address this p...
متن کاملSafe and Eeective Importance Sampling
We present two improvements on the technique of importance sampling. First we show that importance sampling from a mixture of densities, using those densities as control variates, results in a useful upper bound on the asymptotic variance. That bound is a small multiple of the asymptotic variance of importance sampling from the best single component density. This allows one to beneet from the g...
متن کامل