Preface to the Special Cluster on Stochastic Local Search: Recent Developments and Trends

Stochastic local search (SLS) algorithms are among the most powerful techniques for tackling computationally hard problems in various areas of computing science, operations research, business administration, and engineering. SLS range from simple constructive iterative improvement to general-purpose methods, also known as metaheuristics. In recent years, it has become evident that development is complex, typically combines aspects algorithm design implementation with empirical analysis problem-specific background knowledge. The difficulty this process part due complexity being tackled, large number degrees freedom researchers practitioners face when developing algorithms. This needs be supported by a sound methodology addresses issues arising phases design, implementation, tuning, experimental evaluation. addition, more research required understand which best suited particular problem types, gain better insights into relationships between components, parameter settings, characteristics, performance. 2019 workshop given opportunity interested principles practice stochastic meet, present their latest discuss current developments applications. special cluster papers attracted submissions numerous who attended workshop, but contributions other advances area algorithms; these submissions, we selected six. first article, written Vincent Hénaux, Adrien Goëffon, Frédéric Saubion, titled “From fitness landscapes evolution automatic generation.” It proposes an approach based on evolving function automatically generate tackles discrete optimization exploring alternative space. Experiments conducted NK-landscapes; experiments demonstrate ability find landscapes, conducive effective than reference landscape. Sara Tari, Matthieu Basseur, Goëffon study “partial neighbourhood searches” define partial neighborhood (PNLS) technique, where moves random subsets solution. PLNS techniques, balance intensification diversification essentially defined λ $\lambda$ , defines subset size. They analyze three PNLS variants, applying them four combinatorial problems, perform two computational budgets. provide way determine good settings algorithms, some classical variants iterated tabu search. Federico Pagnozzi Thomas Stützle evaluation impact grammar design. such systems, composed components well used describe how may create algorithm. instantiate hybridize different metaheuristics templates, leads configuration space challenging explore effectively. Here, they investigate what happens if permit hybridization at two, one, or zero times. show widely studied permutation flow-shop system only uses doing so provides performance improvements. “Pattern mining-based pruning strategies searches scheduling problems” Arnaud Laurent, Damien Lamy, Benjamin Dalmas Clerc module knowledge extraction extension independent can plugged methods rely runs aim prune elements do not result high-quality solutions. Example results speed up convergence optima same method applied without techniques. article “Two heuristics label printing problem,” Alonso-Pecina, Javier Arellano-Verdejo, Rocio Diego-Celis propose solve using metaheuristics: threshold accepting search, each neighborhoods. proposed improve obtained instances literature. successfully real-world industrial setting. Salim Haddadi discusses “Exponential consecutive block minimization.” A size exponential instance size, called minimization problem. Results benchmarks literature faster gives Last least, would like thank Editor supporting issue, many people Wiley helped us production phase, referees select articles finally authors, submitted work ITOR.