Engineering stochastic local search algorithms

En savoir plus sur le livre

Stochastic local search (SLS) algorithms are widely recognized for their effectiveness in solving complex decision and optimization problems across computer science, operations research, and engineering. Their popularity stems from the conceptual simplicity of many SLS methods and their strong performance on a diverse array of problems, from abstract academic challenges to specific real-world applications. SLS methods include straightforward construction procedures and iterative improvement algorithms, as well as more complex general-purpose schemes known as metaheuristics, such as ant colony optimization, evolutionary computation, iterated local search, memetic algorithms, simulated annealing, tabu search, and variable neighborhood search. Historically, the development of effective SLS algorithms has relied heavily on experience and intuition, resembling an art form. However, recent insights reveal that this development is a complex engineering process that integrates algorithm design, empirical analysis, and problem-specific knowledge across various disciplines, including computer science, operations research, artificial intelligence, and statistics. This process necessitates a robust methodology to address challenges in algorithm design, implementation, tuning, and experimental evaluation.