Hamadi Y., Monfroy E., Saubion F. (eds.). Autonomous Search

Springer, 2012. — 308 p. — ISBN 978-3-642-21433-2, e-ISBN 978-3-642-21434-9.In the foreword of his seminal book on artificial intelligence and on problem solving by man and machine, Jean-Louis Laurière pointed out that one of the main goals of research was to understand how a system (man or machine) may learn, analyze knowledge, transpose and generalize it in order to face real situations and finally solve problems. Therefore, the ultimate goal of this pioneer AI scientist was to build systems that were able to autonomously solve problems, and to acquire and even discover new knowledge, which could be reused later.
Finding the most suitable algorithm and its correct setting for solving a given problem is the holy grail of many computer science researchers and practitioners. This problem was investigated many years ago as the algorithm selection problem. The proposed abstract model suggested extracting features in order to characterize the problem, searching for a suitable algorithm in the space of available algorithms and then evaluating its performance with respect to a set of measures. These considerations are still valid and can indeed be considered at least from two complementary points of view:
Selecting solving techniques or algorithms from a set of possible available techniques;
Tuning an algorithm with respect to a given instance of a problem.
To address these issues, existing works often include tools from different computer science and mathematics areas, especially from machine learning and statistics. Moreover, many approaches have been developed to answer the algorithm selection problem in various fields, as described in the survey of K. Smith-Miles. This present book focuses on the restriction of this general question to the fields of constraint satisfaction, constraint solving, and optimization problems.An Introduction to Autonomous Search
Off-line Configuration
Evolutionary Algorithm Parameters and Methods to Tune Them
Automated Algorithm Configuration and Parameter Tuning
Case-Based Reasoning for Autonomous Constraint Solving
Learning a Mixture of Search Heuristics
On-line Control
An Investigation of Reinforcement Learning for Reactive Search Optimization
Adaptive Operator Selection and Management in Evolutionary Algorithms
Parameter Adaptation in Ant Colony Optimization
New Directions and Applications
Continuous Search in Constraint Programming
Control-Based Clause Sharing in Parallel SAT Solving
Learning Feature-Based Heuristic Functions