Analysing Stochastic Search Heuristics Operating on a Fixed Budget

Abstract

When stochastic search heuristics are used for optimisation they are often stopped after some time has passed and the best search point they have found at this point is used as the solution. Fixed-budget analysis is an analytical perspective that delivers results about the expected quality of the solution in this situation. It allows the comparison of different stochastic search heuristics when only a fixed computational budget is available and it offers a very different perspective from runtime analysis. This chapter introduces and motivates this approach to the theoretical analysis of stochastic search heuristics. It provides basic results, describes a general technique to derive such results from runtime results, covers analytical methods that have been applied and describes a range of different results that have been obtained so far.