Current browse context:
math.OC
Change to browse by:
References & Citations
Mathematics > Optimization and Control
Title: Black-box optimization on hyper-rectangle using Recursive Modified Pattern Search and application to ROC-based Classification Problem
(Submitted on 28 Apr 2016 (v1), revised 20 Sep 2019 (this version, v3), latest version 13 Sep 2023 (v4))
Abstract: In Statistics, multi-modal and non-smooth likelihood (or, objective function) maximization problems often arise with known upper and lower bound of the parameters. A novel derivative-free global optimization technique is developed to optimize any black-box function on a hyper-rectangular euclidean space. In literature, pattern search technique has been shown to be a powerful tool for blackbox optimization. The proposed algorithm follows the principle of pattern search technique where new updated solution is obtained from the current solution making movements (within the constrained sample space) along the coordinates. Before making a jump from the current solution point to a new solution point, objective function is evaluated in several neighborhood points around the current solution and the best solution point is chosen based on the objective function values at those points. Parallel threading can be used to make the algorithm more scalable. Performance of the proposed method is evaluated based on optimization of upto 5000 dimensional multi-modal benchmark functions. The proposed algorithm is shown to perform upto 40 and 368 times faster compared to Genetic Algorithm (GA) and Simulated Annealing (SA) respectively. The proposed method is used to estimate the optimal biomarker combination from Alzheimer data by maximizing the empirical estimates of area under ROC curve.
Submission history
From: Priyam Das [view email][v1] Thu, 28 Apr 2016 20:59:31 GMT (15kb)
[v2] Sat, 17 Dec 2016 05:43:53 GMT (226kb,D)
[v3] Fri, 20 Sep 2019 05:49:22 GMT (898kb,D)
[v4] Wed, 13 Sep 2023 00:18:38 GMT (556kb,D)
Link back to: arXiv, form interface, contact.