Intelligent Monte Carlo Sampling for Decision Making Under Uncertainty

C.-H. Chen (USA) and E. Yücesan (France)


Monte Carlo Simulation, Simulation Uncertainty, Simulation Optimization, Discrete Event Simulation


The process of decision making is usually modeled as a design or optimization problem. Stochastic simulation technology, such as discrete-event simulation and Monte Carlo simulation, has matured over the past decade and is now commonly used to evaluate large-scale real systems with complex stochastic behavior. However, the added flexibility often creates models that are computationally intractable. We present a highly efficient procedure to identify the best design out of k (simulated) competing designs. The objective is to maximize the simulation efficiency, expressed as the probability of correct selection within a given computing budget. Our procedure allocates replications in a way that optimally improves an asymptotic approximation to the probability of correct selection. Numerical testing shows that our approach is much more efficient than all compared methods. Comparisons with the crude ordinal optimization show that our approach can achieve a speedup factor of 3~4 for a 10-design example.

Important Links:

Go Back