Chapter 8. Experiments

Introduction

Experimentation plays an important role in the Algorithm Engineering cycle. It is a powerful tool that amends the traditional and established theoretical methods of algorithm research. Instead of just analyzing the theoretical properties, experiments allow for estimating the practical performance of algorithms in more realistic settings. In other fields related to Computer Science, like for instance Mathematical Programming or Operations Research, experiments have been an indispensable method from the very beginning. Moreover, the results of systematic experimentation may yield new theoretical insights that can be used as a starting point for the next iteration of the whole Algorithm Engineering cycle.

Thereby, a successful experiment is based on extensive planning, an accurate selection of test instances, a careful setup and execution of the experiment, and finally a rigorous analysis and concise presentation of the results. We discuss these issues in this chapter.