Attribute-based Bloom filter tabu search for the Flexible Flowshop Problem

Fazekas, Levente Áron and Nehéz, Károly (2026) Attribute-based Bloom filter tabu search for the Flexible Flowshop Problem. ANNALES MATHEMATICAE ET INFORMATICAE, 63. pp. 55-64. ISSN 1787-6117

Preview

Text 55_64.pdf - Published Version Download (1MB) | Preview

Abstract

In realistic flexible flowshop scheduling, candidate solutions are evaluated through detailed simulation, making optimization fundamentally evaluation-limited. In this regime, classical tabu search becomes increasingly inefficient: exact memory structures grow over time, yet fail to prevent near-duplicate evaluations that induce almost identical simulation trajectories. This paper reframes tabu memory as a constant-time rejection mechanism rather than an exact history structure. We first replace exact tabu memory with Bloom filters, yielding a fixed-size, negligible-cost filtering layer. We then extend this idea to attribute-based tabu, where multiple Bloom filters operate on structural features of solutions rather than their full representations. The resulting multi-filter mechanism enables early rejection of both revisits and structurally similar candidates before simulation. Experimental results show that exact Bloom tabu closely reproduces classical behavior, while attribute-based filtering reduces the number of costly evaluations by over 20% on average. The results highlight a central design principle: in simulation-dominated optimisation, efficiency is achieved not by storing more history, but by making evaluation rare.

Actions (login required)

Edit Item