NDAK16000U Algorithm Engineering (AE)

Algorithm engineering is a discipline between algorithm theory and computing practice. Theoretical algorithmics supplies us with a rich set of algorithms and data structures that, in principle, enable us to solve complex and hard real-world problems. Often the algorithms are designed having the classical random-access machine in mind and the resource requirements of the developed algorithms are analysed in the worst-case or average-case scenario.

In algorithm engineering we design and analyse algorithms for more realistic machine models that take into account the existence of branch predictors, caches, disks, multi-cores, and clusters. In our analysis we take into account the constant factors in the leading terms of the resource bounds. We treat programs as first-class citizens and investigate how algorithms can be turned into reliable and efficient implementations and how these programs can be packaged into easy-to-use software libraries. We do experiments with real-world data and investigate how to solve typical problem instances efficiently.

To summarize, algorithm engineering can be seen as a general methodology for algorithmics. Its heart is an interwoven cycle of design, analysis, implementation, and experimentation. We will design algorithms and prove theorems about them, we will implement our algorithms and do experiments with the implementations, and we will learn best practices of experimentation and library design.

Table of contents

- introduction
- modelling real applications
- realistic models of computation
- algorithm design hierarchy
- meticulous analysis
- implementation aspects
- experimentation
- library design
- case studies