Название: Design of Heuristic Algorithms for Hard Optimization: With Python Codes for the Travelling Salesman Problem
Автор: Eric D. Taillard
Издательство: Springer
Год: 2023
Страниц: 293
Язык: английский
Формат: pdf (true)
Размер: 14.7 MB

This book demonstrates all the steps required to design heuristic algorithms for difficult optimization. The classic problem of the travelling salesman is used as a common thread to illustrate all the techniques discussed. This problem is ideal for introducing readers to the subject because it is very intuitive and its solutions can be graphically represented. The book features a wealth of illustrations that allow the concepts to be understood at a glance.

The book approaches the main metaheuristics from a new angle, deconstructing them into a few key concepts presented in separate chapters: construction, improvement, decomposition, randomization and learning methods. Each metaheuristic can then be presented in simplified form as a combination of these concepts. This approach avoids giving the impression that metaheuristics is a non-formal discipline, a kind of cloud sculpture. Moreover, it provides concrete applications of the travelling salesman problem, which illustrate in just a few lines of code how to design a new heuristic and remove all ambiguities left by a general framework. Two chapters reviewing the basics of combinatorial optimization and complexity theory make the book self-contained. As such, even readers with a very limited background in the field will be able to follow all the content.

One of the reasons for the popularity of metaheuristics is that they allow addressing difficult problems with simple codes. This book contains several pieces of code illustrating how to implement the basic methods discussed. Due to a certain level of abstraction, these principles could be perceived as a sculpture on clouds. The codes eliminate all ambiguity on the inevitable interpretations that can be done when presenting a metaheuristic framework. The computer scientist wishing to develop a heuristic method for a particular problem can be inspired by these codes.

As a source code is useful only when one wants to know all the details of a method, but that it is of little interest when reading. So, we have simplified and shortened the codes, trying to limit them to a single page. These codes come in addition to the text and not the opposite. The reader with little interest in programming or not familiar with the programming language used (from C to Python) can skip them.