Enhancing Stochastic Search Performance by Value-Biased Randomization of Heuristics

Abstract

This paper investigates the utility of introducing randomization as a means of boosting the performance of search heuristics. We introduce a particular approach to randomization, called Value-biased stochastic sampling (VBSS), which emphasizes the use of heuristic value in determining stochastic bias. We offer an empirical study of the performance of value-biased and rank-biased approaches to randomizing search heuristics. We also consider the use of these stochastic sampling techniques in conjunction with local hill-climbing. Finally, we contrast the performance of stochastic sampling search with more systematic search procedures as a means of amplifying the performance of search heuristics.

Simulated Annealing as an Intensification Component in Hybrid Population-based Metaheuristics

“…The literature about applications of metaheuristics to scheduling is quite extended. In (Liao & Juan, 2007) an ACO algorithm for the STWTSDS is proposed, which is able to improve about 86% of the best known results for the Cicirello's benchmark previously found by stochastic search procedures in (Cicirello & Smith, 2005) Recently the Cicirello's best known solutions have been further independently improved in (Cicirello, 2006) by means of a GA approach, in (Lin & Ying, 2006) with three SA, GA and TS algorithms, in (Anghinolfi & Paolucci, 2008) using an ACO approach and in (Anghinolfi & Paolucci, 2007b) with PSO.…”

“…However, constructive heuristics, even if requiring smaller computational efforts, are generally outperformed by improvement, i.e., local search, and metaheuristics approaches. The effectiveness of stochastic search procedures for the STWTSDS is shown in (Cicirello & Smith, 2005), where the authors compare a value-biased stochastic sampling (VBSS), a VBSS with hill-climbing (VBSS-HC) and a simulated annealing (SA), to limited discrepancy search (LDS) and heuristic-biased stochastic sampling (HBSS) on a 120 benchmark problem instances for the STWTSDS problem defined by Cicirello in (Cicirello, 2003). The literature about applications of metaheuristics to scheduling is quite extended.…”

“…Note that this benchmark was used for testing various metaheuristic approaches recently appeared in the literature as (Cicirello & Smith, 2005), (Liao & Juan, 2007), (Cicirello, 2006), (Lin & Ying, 2006), (Anghinolfi & Paolucci, 2008) and (Anghinolfi & Paolucci, 2007b). The benchmark was produced by generating 10 instances for each combination of three different factors usually referenced in the literature (for a definition and discussion see, e.g., (Pinedo, 1995)): the due date tightness δ, the due date range R, and the setup time severity ξ, selected as follows: δ∈{0.3, 0.6, 0.9}, R∈{0.25, 0.75}, ξ∈{0.25, 0.75}.…”

Genetic Algorithm Parameter Control: Application to Scheduling with Sequence-Dependent Setups

“…1 This benchmark set has been used by many researchers for a variety of search algorithms, such as dynamic programming [34], neighborhood search [24], iterated local search [39], valuebiased stochastic sampling [14], genetic algorithms [7], simulated annealing [8], ant colony optimization [23], etc.…”

A hybrid cuckoo search metaheuristic algorithm for solving single machine total weighted tardiness scheduling problems with sequence dependent setup times

“…In order to evaluate the performance of the proposed algorithm, the benchmark problems studied by Cicirello and Smith (2005) are also considered in this paper. The benchmark problem consists of 120 problems each with 60 jobs.…”