23 January 2018 in IEEE Transactions on Automatic Control
Under a Bayesian framework, we formulate the fully sequential sampling and selection decision in statistical ranking and selection as a stochastic control problem, and derive the associated Bellman equation. Using value function approximation, we derive an approximately optimal allocation policy. We show that this policy is not only computationally efficient but also possesses both one-step-ahead and asymptotic optimality for independent normal sampling distributions. Moreover, the proposed allocation policy is easily generalizable in the approximate dynamic programming paradigm.
18 July 2016 in IEEE Transactions on Evolutionary Computation
Particle swarm optimization (PSO) is a popular metaheuristic for deterministic optimization. Originated in the interpretations of the movement of individuals in a bird flock or fish school, PSO introduces the concept of personal best and global best to simulate the pattern of searching for food by flocking and successfully translate the natural phenomena to the optimization of complex functions. Many real-life applications of PSO cope with stochastic problems. To solve a stochastic problem using PSO, a straightforward approach is to equally allocate computational effort among all particles and obtain the same number of samples of fitness values. This is not an efficient use of computational budget and leaves considerable room for improvement. This paper proposes a seamless integration of the concept of optimal computing budget allocation into PSO to improve the computational efficiency of PSO for stochastic optimization problems. We derive an asymptotically optimal allocation rule to intelligently determine the number of samples for all particles such that the PSO algorithm can efficiently select the personal best and global best when there is stochastic estimation noise in fitness values. We also propose an easy-to-implement sequential procedure. Numerical tests show that our new approach can obtain much better results using the same amount of computational effort.
1 January 2024 in Manufacturing & Service Operations Management
Problem definition: Personalized medicine (PM) seeks the best treatment for each patient among a set of available treatment methods. Because a specific treatment does not work well on all patients, traditionally, the best treatment was selected based on the doctor’s personal experience and expertise, which is subject to human errors. In the meantime, stochastic models have been well developed in the literature for a lot of major diseases. This gives rise to a simulation-based solution for PM, which uses the simulation tool to evaluate the performance for pairs of treatment and patient biometric characteristics and, based on that, selects the best treatment for each patient characteristic. Methodology/results: In this research, we extend the ranking and selection (R&S) model in simulation-based decision making to solving PM. The biometric characteristics of a patient are treated as a context for R&S, and we call it contextual ranking and selection (CR&S). We consider two formulations of CR&S with small and large context spaces, respectively, and develop new techniques for solving them and identifying the rate-optimal budget allocation rules. Based on them, two selection algorithms are proposed, which can be shown to be numerically superior via a set of tests on abstract and real-world examples. Managerial implications: This research provides a systematic way of conducting simulation-based decision-making for PM. To improve the overall decision quality for the possible contexts, more simulation efforts should be devoted to contexts in which it is difficult to distinguish between the best treatment and non-best treatments, and our results quantify the optimal trade-off of the simulation efforts between the pairs of contexts and treatments. Funding: J. Du is partially supported by the National Natural Science Foundation of China [Grant 72091211]. C.-H. Chen is partially supported by the National Science Foundation under Awards FAIN212368. Supplemental Material: The e-companion is available at https://doi.org/10.1287/msom.2022.0232 .
4 September 2012 in IEEE Transactions on Automatic Control
Common random numbers and the standard clock method are examples of effective variance reduction techniques that also share information and simulation resources when generating realizations of different simulated systems whose performances are being compared. This sharing of computing resources and the potentially widely different computational requirements for different simulation models are important considerations in allocating simulation replications among the candidate designs with the objective of maximizing the probability of selecting the best design, and we formulate the optimal computing budget allocation problem under this scenario. The resulting formulation leads to an optimization problem that can be viewed as a generalization of a correlated version considered in earlier work. An approximation to the problem is introduced to allow a tractable solution, for which a heuristic two-stage sequential allocation algorithm is proposed, and several numerical examples are used to illustrate the potential improvements that can be gained.
10 December 2018 in IEEE Transactions on Automatic Control
Simulation is often used to estimate the performance of alternative system designs for selecting the best. For a complex system, high-fidelity simulation is usually time-consuming and expensive. In this paper, we provide a new framework that integrates information from the multi-fidelity models to increase efficiency for selecting the best. A Gaussian mixture model is introduced to capture performance clustering information in the multi-fidelity models. Posterior information obtained by a clustering analysis incorporates both cluster-wise information and idiosyncratic information for each design. We propose a new budget allocation method to efficiently allocate high-fidelity simulation replications, utilizing posterior information. Numerical experiments show that the proposed multi-fidelity framework achieves a significant boost in efficiency.