BFS 2002

Contributed Talk

Robbins-Monro Algorithm, Variance Reduction Technique

Bouhari Arouna

We present in our paper an original variance reduction technique for Monte Carlo methods.
By an elementary version of Girsanov theorem, we introduce a drift term in the price computation. Afterwards our main idea is to use a truncated version of the Robbins-Monro algorithm to find the optimal drift that reduce variance. We proved that for a large class of payoff functions, this version of the RM algorithm converges a.s. to the optimal drift. Then we illustrate the method by an application to some options pricing.