Adaptive tabu search with strategic oscillation for the bipartite boolean quadratic programming problem with partitioned variables

•Propose an adaptive tabu search with strategic oscillation approach for BQP-PV.•Combine different move operators to conduct neighborhood exploration.•Design an adaptive tabu tenure management strategy by utilizing the search history.•Employ frequency-driven strategic oscillation to diversify the se...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Information sciences Ročník 450; s. 284 - 300
Hlavní autoři: Wang, Yang, Wu, Qinghua, Punnen, Abraham P., Glover, Fred
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Inc 01.06.2018
Témata:
Adaptive tabu search
Heuristic
Quadratic assignment
Strategic oscillation
Strategic oscillation
Quadratic assignment
Adaptive tabu search
Heuristic
ISSN:0020-0255, 1872-6291
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•Propose an adaptive tabu search with strategic oscillation approach for BQP-PV.•Combine different move operators to conduct neighborhood exploration.•Design an adaptive tabu tenure management strategy by utilizing the search history.•Employ frequency-driven strategic oscillation to diversify the search.•Find improved solutions for 14 instances. The bipartite boolean quadratic programming problem with partitioned variables (BQP-PV) is an NP-hard combinatorial optimization problem that accommodates a variety of real-life applications. We propose an adaptive tabu search with strategic oscillation (ATS-SO) approach for BQP-PV, which employs a multi-pass search framework where each pass consists of an initial constructive phase, an adaptive tabu search phase and a frequency-driven strategic oscillation phase. In particular, the adaptive tabu search phase combines different move operators to collectively conduct neighborhood exploration and an adaptive tabu tenure management mechanism that obviates the task of determining a proper tabu tenure. The frequency-driven strategic oscillation phase diversifies the search when the search reaches a critical solution, drawing on a destructive procedure to unassign some variables by reference to frequency memory and a constructive procedure to re-assign these variables utilizing both frequency memory and problem specific knowledge. Computational experiments on five classes of problem instances indicate that the proposed ATS-SO algorithm is able to find improved solutions for 14 instances and match the best known solutions for all remaining instances, whereas no previous method has succeeded in finding the previous best solutions for all instances. Statistical tests indicate that ATS-SO significantly outperforms the state-of-the-art algorithms in the literature.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2018.03.045