Theoretical and numerical study of the performance of `Pincus' optimisation method (Q6108480)

Summary: In this work, we are interested in the numerical and theoretical study of the so called `Pincus algorithm', which is a stochastic optimisation method. It is based on a representation of the optimum as a limit of a ratio of two expectations, computed using a Monte Carlo approximation. First, we theoretically study the convergence of Pincus algorithm. Then, numerical computations are performed: we show the algorithm advantages and limits, in comparison with other methods in terms of robustness, speed and dependence on the dimension.