Professional Documents
Culture Documents
kevin.tierney@uni-bielefeld.de
Professor for Decision and Operation Technologies
Bielefeld University
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International license.
Outline
Outline
Reminder:
I Goal: Solve optimization problems
This video:
I Heuristics vs. Metaheuristics
I Classes of Heuristics
I Metaheuristics covered in this class
Heuristic methods
Definition
Heuristic methods attempt to solve problems without providing guarantees of
optimality or solution quality using tricks or “rules of thumb”.
What to consider. . .
1. How much time do you have to compute a solution?
2. What is the value of optimality?
3. How accurate (realistic) is your model?
Types of heuristics
Construction heuristics:
I Iteratively builds a solution
I (Usually) decomposes a solution into solution components
Neighborhood heuristics:
I Provides one or more solutions based on a given solution
I Similar solutions = “neighborhood” of a solution
Classes of heuristics
Metaheuristics
Definition 2
“[M]etaheuristics are ’solution methods that orchestrate an interaction between local
improvement procedures and higher level strategies to create a process capable of
escaping from local optima and performing a robust search of a solution space.’ ”
– Michel Gendreau and Jean-Yves Potvin, Handbook of Metaheuristics, Second Edition
Overview of Metaheuristics
Heuristics Metaheuristics
I Problem specific I General, problem independent
I Find good solutions, but can get I Built in strategies to avoid local
stuck in a local optimum optima (can and will get stuck
I Often focus on solving a particular anyway)
part or structure of a problem I “High-level” methods for selecting
and applying heuristics
Development of metaheuristics
Summary
In this video. . .
I We looked at the different types of heuristics and metaheuristics
I We compared and contrasted heuristics and metaheuristics
I We learned an important fact about heuristics and metaheuristics: namely, that
they are not guaranteed to find optimal solutions.