Design of Modern Heuristics: Principles and Application by Franz Rothlauf

By Franz Rothlauf

Most textbooks on glossy heuristics give you the reader with certain descriptions of the performance of unmarried examples like genetic algorithms, genetic programming, tabu seek, simulated annealing, and others, yet fail to educate the underlying suggestions at the back of those diversified ways.

The writer takes a special procedure during this textbook by means of targeting the clients' wishes and answering 3 primary questions: First, he tells us which difficulties sleek heuristics are anticipated to accomplish good on, and which might be left to standard optimization equipment. moment, he teaches us to systematically layout the "right" sleek heuristic for a specific challenge via supplying a coherent view on layout parts and dealing rules. 3rd, he exhibits how we will utilize problem-specific wisdom for the layout of effective and powerful sleek heuristics that resolve not just small toy difficulties but in addition practice good on huge real-world difficulties.

This ebook is written in an easy-to-read variety and it truly is geared toward scholars and practitioners in laptop technological know-how, operations learn and data structures who are looking to comprehend sleek heuristics and have an interest in a advisor to their systematic layout and use.

Show description

Read or Download Design of Modern Heuristics: Principles and Application PDF

Similar linear programming books

The Stability of Matter: From Atoms to Stars

During this assortment the reader will locate common effects including deep insights into quantum structures mixed with papers at the constitution of atoms and molecules, the thermodynamic restrict, and stellar constructions.

Generalized Linear Models, Second Edition (Chapman & Hall CRC Monographs on Statistics & Applied Probability)

The luck of the 1st variation of Generalized Linear versions resulted in the up-to-date moment variation, which maintains to supply a definitive unified, therapy of tools for the research of numerous varieties of facts. at the present time, it continues to be renowned for its readability, richness of content material and direct relevance to agricultural, organic, wellbeing and fitness, engineering, and different functions.

Switched Linear Systems: Control and Design (Communications and Control Engineering)

Switched linear platforms have loved a selected progress in curiosity because the Nineties. the massive volume of knowledge and ideas therefore generated have, before, lacked a co-ordinating framework to concentration them successfully on a number of the primary concerns comparable to the issues of sturdy stabilizing switching layout, suggestions stabilization and optimum switching.

AMPL: A Modeling Language for Mathematical Programming

AMPL is a language for large-scale optimization and mathematical programming difficulties in construction, distribution, mixing, scheduling, and plenty of different functions. Combining widespread algebraic notation and a strong interactive command atmosphere, AMPL makes it effortless to create types, use a wide selection of solvers, and look at strategies.

Additional resources for Design of Modern Heuristics: Principles and Application

Example text

The second example is the needle-in-a-haystack (NIH) problem. Following its name, the goal is to find a needle in a haystack. In this problem, a metric exists defining distances between solutions, but there is no meaningful relationship between the metric and the objective value (needle found or not) of different solutions. When physically searching in a haystack for a needle, there is no good strategy for choosing promising areas of the haystack that should be searched in the next search step.

Two solutions x and y are denoted as neighbors if d(x, y) = dmin . Often, d can be normalized to dmin = 1. A fitness landscape can be described using a graph GL with a vertex set V = X and an edge set E = {(x, y) ∈ X × X | d(x, y) = dmin } (Reeves, 1999a; Merz and Freisleben, 2000b). The objective function assigns an objective value to each vertex. We assume that each solution has at least one neighbor and the resulting graph is connected. Therefore, an edge exists between neighboring solutions.

If we know that an algorithm can solve a problem, we automatically have an upper bound on the difficulty of the problem, which is just the complexity of the algorithm. For example, we study the problem of finding a friend’s telephone number in the telephone book. The most straightforward approach is to search through the whole book starting from “A”. The effort for doing this increases linearly with the number of names in the book. Therefore, we have an upper bound 24 2 Optimization Problems on the difficulty of the problem (problem has at most linear complexity) as we know a linear algorithm that can solve the problem.

Download PDF sample

Rated 4.99 of 5 – based on 15 votes