User Manual

1. Introduction

1.1 Target Audience

This manual is intended for end users, domain experts and students who are not necessarily familiar with algorithm development in .NET. It therefore focuses on the graphical user interface and how to configure, customize and extend the available standard algorithm already available in HeuristicLab 3.3 for a particular problem without writing (a lot of) code. We believe that a successive transfer of competence in algorithm development from heuristic optimization experts to users working on real-world applications would be very beneficial for the community as a whole. This manual will slowly guide you through a number of tutorials after which we hope even novice users will no longer have to use metaheuristics as black box techniques, but can use them as algorithms which can be modified and easily tuned to specific problem situations.

1.2 HeuristicLab Software Architecture

The core of HeuristicLab is its algorithm model. It is not necessary to understand more than an abstract level of the core language as an end user. We provide a short summary of the most important components here:

Core concepts:

• Data: All HeuristicLab 3.3 data objects can be saved, restored and viewed. Standard data types such as integers, doubles, strings, or arrays that do not offer these properties are wrapped in HeuristicLab 3.3 objects.
• Variables: Data objects are linked to a name by storing them in a variable.
• Scopes: Scopes are containers for variables. To access a variable in a scope, the variable name is used as an identifer (has to be unique per scope). Scopes can be hierarchically organized as trees (subscope hierarchy).
• Operators: Each algorithm is a sequence of instructions, which are called operators in HeuristicLab 3.3. Operators are applied on scopes to access/manipulate variables and subscopes.
• Parameters: Each operator defines parameters for each variable it expects and (possibly) manipulates.
• Engines: Algorithms are represented as operator graphs and executed step-by-step on virtual machines called engines.

We recommend having a look at the Architecture? wiki page to get a more detailed description of the algorithm model. You can also revisit this page later on, as it is not really required for the introduction section.

2. Getting Started

If you start HeuristicLab, a splash screen will appear while all the plugins are loaded. Afterwards you can select a HeuristicLab application on the start screen:

The standard configuration, which you can download on this homepage, is shipped with two applications:

• HeuristicLab Optimizer: The optimizer is the core application of the HeuristicLab Framework, providing a variety of metaheuristic algorithms, standard problem implementations, a comfortable test bench and modules for algorithm development and visualization. As an end user you will mostly work with the Optimizer.
• HeuristicLab Plugin Manager: The HeuristicLab architecture is based on plugins. All optimization algorithms and problems are realized as plugins and interact with each other via well-defined and extensible interfaces. The plugin manager allows you to browse, install and update available plugins from our centralized update location or manage your own plugins.

3. HeuristicLab Optimizer

3.1 Start Working

Follow these steps to start working with HeuristicLab Optimizer:

1. Open an algorithm
   • click New Item in the toolbar and select an algorithm or click Open File in the toolbar and load an algorithm from a file
2. Open a problem in the algorithm
   • in the Problem tab of the algorithm click New Problem and select a problem or click Open Problem and load a problem from a file
3. Set parameters
   • set problem parameters in the Problem tab of the algorithm
   • set algorithm parameters in the Parameters tab of the algorithm
4. Run the algorithm
   • click Start/Resume Algorithm to execute the algorithm (if the button is grayed out some parameters of the algorithm or the problem still have to be set)
   • wait for the algorithm to terminate or click Pause Algorithm to interrupt its execution or click Stop Algorithm to stop its execution
5. Check results
   • check the results on the Results tab of the algorithm
   • click Start/Resume Algorithm to continue the algorithm or click Reset Algorithm to prepare a new run

3.2 Samples

For you convenience we have prepared a set of samples that are shipped with HeuristicLab. These samples provide a set of preconfigured algorithms for particular problems. For example, if you want to solve a TSP problem with genetic algorithms our preconfigured GA for the ch130 problem might be worth a look.

4. Tutorials

Are you looking for predefined algorithms which can be executed immediately or an easy step-by-step guide to the parameterization of a particular algorithm? In this section we will provide tutorials on different aspects of HeuristicLab. Most of them are not yet written, but we strive to catch up with the documentation. Contributions are always welcome!
You can also have a look at this page? for slides of tutorials that we held at conferences.

4.1 Optimizing External Applications

The ​Race Car Setup Optimization Competition was organized for the ​EvoStar 2010 conference. The problem is a classic "simulation-based optimization problem" with a simulator, a simulation model, and several configuration parameters that have to be optimized. The following Tutorial? will show you how to configure an external evaluation problem for that particular problem.

5. Plugin Infrastructure

The Plugin Manager Console? allows you to install/remove plugins and upgrade your plugins when new versions become available.

6. Problems

The following problems are currently available in HeuristicLab.

• Artificial Ant Problem?
• External Evaluation Problem?
• Knapsack Problem?
• OneMax Problem
• Single Objective Test Function
• Symbolic Regression Problem
• Symbolic Classification Problem
• Travelling Salesman Problem?
• Vehicle Routing Problem?
• Quadratic Assignment Problem
• Scheduling
• Clustering
• User-defined Problem?

7. Algorithms

• Evolution Strategy?
• Genetic Algorithm?
• Island Genetic Algorithm
• Island Offspring Selection Genetic Algorithm
• Local Search
• Non-dominated Sorting Genetic Algorithm?
• Offspring Selection Genetic Algorithm?
• Particle Swarm Optimization?
• SASEGASA
• Simulated Annealing
• Tabu Search
• User-Defined Algorithm
• Variable Neighborhood Search
• Linear Regression
• Linear Discriminant Analysis
• Support Vector Machine
• k-Means
• User-defined Algorithm

8. Operators

Each algorithm consists of a sequence of clearly defined, unambiguous and executable instructions. We call these atomic building blocks operators. A list of the most important operators is available in the wiki and we try to update it frequently. If the operator you are looking for is not documented there we recommend having a look at the description of the operator in the Optimizer. The description is usually sufficient to apply the operator successfully.

9. Testing and Analysis

• Batch Run?: A run in which an algorithm is executed a given number of times (repetitions).
• Experiment: An experiment which contains multiple batch runs of algorithms.

10. Feedback

Any feedback, questions, problems or requests for new features? Please have a look at our support & feedback? wiki page to see how to contact the HeuristicLab development team.