Direct Search
Pattern search solver for derivative-free optimization,
constrained or unconstrained
Direct search is an efficient algorithm for solving smooth or nonsmooth
optimization problems. Try patternsearch
first for most
nonsmooth problems.
Functions
Live Editor Tasks
Optimize | Optimize or solve equations in the Live Editor (Since R2020b) |
Topics
Problem-Based Direct Search
- Optimize Nonsmooth Function Using patternsearch, Problem-Based
Basic example minimizing a nonsmooth function in the problem-based approach. - Constrained Minimization Using Pattern Search, Problem-Based
Usepatternsearch
to minimize an objective function subject to bounds and nonlinear constraints. - Effects of Pattern Search Options, Problem-Based
Visualize and tune direct search in the problem-based approach. - Search and Poll, Problem-Based
Examples showing the utility of search in addition to poll methods in the problem-based approach.
Solver-Based Direct Search Basics
- Optimize Using the GPS Algorithm
Provides an example of solving an optimization problem using pattern search. - Coding and Minimizing an Objective Function Using Pattern Search
Shows how to write an objective function including extra parameters or vectorization. - Constrained Minimization Using patternsearch and Optimize Live Editor Task
Example using linear constraints and nonlinear constraints inpatternsearch
. - Explore patternsearch Algorithms
This example shows the effect of choosing differentpatternsearch
algorithms. - Explore patternsearch Algorithms in Optimize Live Editor Task
This example shows the effect of choosing differentpatternsearch
algorithms using the Optimize Live Editor task. - Constrained Minimization Using Pattern Search, Solver-Based
Use constraints in direct search. - Effects of Pattern Search Options
Visualize and tune direct search. - Set Options
Shows how to set and examine options forpatternsearch
. - Optimization of Stochastic Objective Function
Pattern search can minimize a function even in the presence of noise. - Search and Poll
Examples showing the utility of search in addition to poll methods.
Solver-Based Specialized Tasks
- Polling Types
Examines the effects of polling options, including theUseCompletePoll
option. - Set Mesh Options
Examines the effect of different mesh expansion and contraction factors. - Custom Plot Function
Shows how to write and use a plot function forpatternsearch
. - Pattern Search Climbs Mount Washington
Shows the stepspatternsearch
takes by using custom plot functions. - Optimization of Stochastic Objective Function
Pattern search can minimize a function even in the presence of noise. - Vectorize the Objective and Constraint Functions
How to gain speed using vectorized function evaluations. - Optimize ODEs in Parallel
Save time by calling an expensive subroutine just once and computing an ODE solution in parallel usingpatternsearch
orga
. - Optimize Simulink Model in Parallel
This example shows how to optimize a Simulink® model in parallel using several Global Optimization Toolbox solvers.
Direct Search Background
- What Is Direct Search?
Introduces direct search and pattern search. - Pattern Search Terminology
Explains some basic pattern search terminology. - How Pattern Search Polling Works
Provides an overview of direct search algorithms. - Nonuniform Pattern Search (NUPS) Algorithm
Description of the NUPS algorithm. - Searching and Polling
Describes how search methods work with polling steps. - Setting Solver Tolerances
Stopping conditions and their associated options. - Nonlinear Constraint Solver Algorithm for Pattern Search
Explains the Augmented Lagrangian Pattern Search (ALPS). - Pattern Search Options
Explore the options for pattern search.