The University of Southampton
University of Southampton Institutional Repository

Parameter optimisation for search heuristics via a barrier tree Markov model

Parameter optimisation for search heuristics via a barrier tree Markov model
Parameter optimisation for search heuristics via a barrier tree Markov model

The quality of solution provided by a search heuristic on a particular problem is by no means an absolute value. Most heuristics are controlled by a set of parameters, upon which the performance is heavily dependent. We construct Markov models for search heuristics on specific problem instances, and model the relationship between the quality of search and the choice of parameters.

For any problem instance of nontrivial size, the state space for this model is large enough so as to make computation infeasible. Our solution is to use a reduced-state model of the search space, by amalgamating many similar points. We use a model based on a level-accessible barrier tree, grouping regions of the search space according to the reachability of minima.

Optimal annealing and mutation schedules are produced, minimising either “where-you-are” or “best-so-far” cost, over binary perception, spin-glass and Max-SAT problems. The predictions of the model are found to be consistently over-optimistic; we discuss reasons for this and suggest some possible refinements to the model.

A population-based variant of simulated annealing is briefly examined, where annealing temperature is adjusted according to performance. We later optimise the average first-passage time for several special-case heuristics, comparing the minimal times across a range of problems.

University of Southampton
Benfold, William
6c2d6619-9e08-4ae6-a739-a99df0abe02a
Benfold, William
6c2d6619-9e08-4ae6-a739-a99df0abe02a

Benfold, William (2007) Parameter optimisation for search heuristics via a barrier tree Markov model. University of Southampton, Doctoral Thesis.

Record type: Thesis (Doctoral)

Abstract

The quality of solution provided by a search heuristic on a particular problem is by no means an absolute value. Most heuristics are controlled by a set of parameters, upon which the performance is heavily dependent. We construct Markov models for search heuristics on specific problem instances, and model the relationship between the quality of search and the choice of parameters.

For any problem instance of nontrivial size, the state space for this model is large enough so as to make computation infeasible. Our solution is to use a reduced-state model of the search space, by amalgamating many similar points. We use a model based on a level-accessible barrier tree, grouping regions of the search space according to the reachability of minima.

Optimal annealing and mutation schedules are produced, minimising either “where-you-are” or “best-so-far” cost, over binary perception, spin-glass and Max-SAT problems. The predictions of the model are found to be consistently over-optimistic; we discuss reasons for this and suggest some possible refinements to the model.

A population-based variant of simulated annealing is briefly examined, where annealing temperature is adjusted according to performance. We later optimise the average first-passage time for several special-case heuristics, comparing the minimal times across a range of problems.

Text
1038773.pdf - Version of Record
Available under License University of Southampton Thesis Licence.
Download (2MB)

More information

Published date: 2007

Identifiers

Local EPrints ID: 466093
URI: http://eprints.soton.ac.uk/id/eprint/466093
PURE UUID: 5c11cdbc-c294-4e8e-80ed-2c4fe0e54f05

Catalogue record

Date deposited: 05 Jul 2022 04:17
Last modified: 16 Mar 2024 20:30

Export record

Contributors

Author: William Benfold

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×