Black-box optimization on hyper-rectangle using recursive modified pattern search and application to ROC-based classification problem
From MaRDI portal
Publication:6062853
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.
Recommendations
- Recursive modified pattern search on high-dimensional simplex: a blackbox optimization technique
- Optimization algorithm based on densification and dynamic canonical descent
- Optimization of black-box problems using Smolyak grids and polynomial approximations
- Finding local optima of high-dimensional functions using direct search methods
- Deterministic global derivative-free optimization of black-box problems with bounded Hessian
Cites work
- A literature survey of benchmark functions for global optimisation problems
- A survey on direct search methods for blackbox optimization and their applications
- A trust region method based on interior point techniques for nonlinear programming.
- Algorithm 909: NOMAD: nonlinear optimization with the MADS algorithm
- An Algorithm for Least-Squares Estimation of Nonlinear Parameters
- Combining diagnostic test results to increase accuracy
- Combining multiple markers for multi-category classification: an ROC surface approach
- Combining Predictors for Classification Using the Area under the Receiver Operating Characteristic Curve
- Constrained derivative-free optimization on thin domains
- Efficient global optimization of expensive black-box functions
- Exact matrix completion via convex optimization
- Global optimization by multilevel coordinate search
- GLODS: global and local optimization using direct search
- scientific article; zbMATH DE number 2002582 (Why is no real title available?)
- scientific article; zbMATH DE number 194544 (Why is no real title available?)
- scientific article; zbMATH DE number 845714 (Why is no real title available?)
- Interior-point methods
- Introduction to Derivative-Free Optimization
- Mesh Adaptive Direct Search Algorithms for Constrained Optimization
- Nonsmooth optimization through mesh adaptive direct search and variable neighborhood search
- On the Convergence of Pattern Search Algorithms
- Optimization by simulated annealing
- Parallel Space Decomposition of the Mesh Adaptive Direct Search Algorithm
- Pattern Search Algorithms for Bound Constrained Minimization
- Recursive modified pattern search on high-dimensional simplex: a blackbox optimization technique
- ROC analysis with multiple classes and multiple tests: methodology and its application in microarray studies
- The Power of Convex Relaxation: Near-Optimal Matrix Completion
- Variable Selection via Nonconcave Penalized Likelihood and its Oracle Properties
Cited in
(2)
This page was built for publication: Black-box optimization on hyper-rectangle using recursive modified pattern search and application to ROC-based classification problem
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6062853)
