Malliavin Calculus with Weak Derivatives for Counterfactual Stochastic Optimization
Authors:
- Vikram KrishnamurthyVKVikram Krishnamurthy
- Luke SnowLSLuke Snow
Preprint
Abstract
We study counterfactual stochastic optimization of conditional loss functionals under misspecified and noisy gradient information. The difficulty is that when the conditioning event has vanishing or zero probability, naive Monte Carlo estimators are prohibitively inefficient; kernel smoothing, though common, suffers from slow convergence. We propose a two-stage kernel-free methodology. First, we show using Malliavin calculus that the conditional loss functional of a diffusion process admits an exact representation as a Skorohod integral, yielding variance comparable to classical Monte-Carlo variance. Second, we establish that a weak derivative estimate of the conditional loss functional with respect to model parameters can be evaluated with constant variance, in contrast to the widely used score function method whose variance grows linearly in the sample path length. Together, these results yield an efficient framework for counterfactual conditional stochastic gradient algorithms in rare-event regimes.Keywords
All Related Versions
This publication has 8 references indexed in Scilit:
- Multikernel Passive Stochastic Gradient Algorithms and Transfer LearningIEEE Transactions on Automatic Control, 2021
- Measure-Valued Differentiation for Markov ChainsJournal of Optimization Theory and Applications, 2007
- Sensitivity Analysis Using Itô--Malliavin Calculus and Martingales, and Application to Stochastic Optimal ControlSIAM Journal on Control and Optimization, 2005
- On the Malliavin approach to Monte Carlo approximation of conditional expectationsFinance and Stochastics, 2004
- Stochastic Approximation and Recursive Algorithms and Applications2003
- Applications of Malliavin calculus to Monte-Carlo methods in finance. IIFinance and Stochastics, 2001
- Optimization of Stochastic Models: The Interface between Simulation and Optimization.Journal of the Operational Research Society, 1998
- Passive stochastic approximation with constant step size and window widthIEEE Transactions on Automatic Control, 1996