Reactive search and intelligent optimization (Q931253)

Reactive Search involves the integration of sub-symbolic machine learning techniques into search heuristics for solving complex optimization problems. Methodologies of interest for Reactive Search include machine learning and statistics, in particular reinforcement learning, active or query learning, neural networks, and meta-heuristics (\url{http://www.reactive-search.org/}). It refers to a class of heuristics that automatically adjust their working parameters during the optimization phase (\url{http://en.wikipedia.org/wiki/Reactive-search}). Intelligent optimization, a superset of Reactive Search, refers to a more extended area of research, including online and offline schemes based on the use of memory, adaptation, incremental development of models, experimental algorithmics applied to optimization, intelligent tuning and design of heuristics (\url{http://www.reactive-search.org/}). In this context, it is worth to mention the LION (machine Learning and Intelligent OptimizatioN) laboratory at the University of Trento (Italy) which fosters research and development in intelligent optimization and reactive search techniques for solving relevant problems arising in different application areas, including intelligent transportation systems, computer networks and mobility, mobile services and ubiquitous computing, social networks, clustering and pattern recognition in bio-informatics (\url{http://www.intelligent-optimization.org/}). All the authors of this book are members of LION. As the authors underline in the beginning of the book, they intended to write a study on learning for problem solving using heuristics and they succeeded in doing that. They describe LION as the intersection of operations research (optimization), computer science and machine learning and neural nets. They aim to present in this book the main principles and some fresh intuition for these approaches. They are mainly focusing on methods than on problems. The architecture of the chapters consists in providing: the basic issues and algorithms, the identification of the critical parameters of each method, and opportunities and schemes for automated tuning of these parameters. Among the main ideas, let us mention: reacting on neighborhood, on annealing schedule, on objective function, supervised learning, reinforcement learning, racing, teams of interacting solvers etc. The book is written in a very `friendly' style, although it presents serious ideas and concepts in the area of heuristics and optimization.