A Unified View of Infinitesimal Perturbation Analysis and Likelihood Ratios
Abstract
A view of the likelihood ratio (LR) gradient estimation technique (also called the score function (SF) method) is presented under which infinitesimal perturbation analysis (IPA) can be viewed as a (degenerate) special case, by selecting appropriately what the random component omega effectively represents. Varying the actual meaning of omega (i.e., defining the underlying sample space in different ways) might define different variants of the LR method, some of them mixing IPA with more traditional LR. We illustrate this by many examples. We also give general conditions under which the gradient estimators are unbiased.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1989
- Accession Number
- ADA210682
Entities
People
- Pierre L'ecuyer
Organizations
- Stanford University