Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.639084
Title: Local search and simulation
Author: So, D. G.
Awarding Body: University College of Swansea
Current Institution: Swansea University
Date of Award: 1995
Availability of Full Text:
Access through EThOS:
Abstract:
Simulation has long been recognised as a powerful technique for analysing complex systems which are mathematically intractable. The objectives of simulation experiments are essentially of two types, namely for investigative and optimisation purposes. As far as the latter is concerned, analysts essentially aim to find the combination of input parameters to the system being investigated so as to optimise some performance measures. The discovery of a 'good' combination of input parameters frequently involves a long and tedious trial and error process which is often computationally demanding. This thesis concerns the automation of the process. In particular, various search algorithms are developed within the framework of local search. These algorithms are used to automatically search for 'promising', or even optimal, parameter settings where the function to be optimised is the output of a simulation model. An important distinction between the work described in this thesis and the more conventional use of local search lies in the nature of the cost function. While the conventional application of local search mainly focuses on problems with a deterministic cost function, this work considers problems in which the cost function is subject to some stochastic infrastructure. To ensure a successful implementation of local search to such problems, it is crucial that the stochastic variation in the cost function is explicitly taken into account. Various strategies to achieve this end are identified and ways in which they can be incorporated into the standard acceptance criteria of local search algorithms are described.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.639084  DOI: Not available
Share: