Suggestions For BSc/MSc Projects

Christine Zarges, CS room 232, c.zarges@cs.bham.ac.uk
Office hours: Tuesday 2-4pm

My research is centred around nature-inspired algorithms, in particular (but not limited to) artificial immune systems and evolutionary algorithms. While evolutionary algorithms are inspired by natural evolution, artificial immune systems are build after the immune system of vertebrates.

I am interested in both, theoretical and practical aspects of such algorithms in different areas of application. I am happy to supervise projects (BSc and MSc) that fall in the broad area of natural computation and computational intelligence. Previous knowledge in this area is an advantage but not a necessity.
In the following you can find a short list of some project ideas. Please feel free to contact me via email if you are interested. You are also welcome to talk to me about your own ideas!

Artificial Immune Systems in Dynamic Environments

Real-world problems are often subject to change over time. In recent years, the field of evolutionary dynamic optimisation has become one of the most active research areas in evolutionary computation. There are also some approaches using artificial immune systems.

The purpose of this project is to investigate different approaches in dynamic optimisation. An emphasis should be on the development of a new artificial immune system and compare it to existing approaches from other areas of research.

Artificial Immune Systems for Classification Problems

One of the most natural tasks of an artificial immune systems is classification. The most prominent examples of such classification algorithms are negative selection and the dendritic cell algorithm. Recently, a first thorough empirical investigation of a new approach of string-based negative selection was presented.

The purpose of this project is to execute a further analysis of immune-inspired classification. In particular, recent developments of the (deterministic) dendritic cell algorithm should be implemented and compared against other approaches.

Evaluation of algorithms for the Vertex Cover Problems

Recent theoretical analyses indicate that the B-Cell Algorithm, a specific artificial immune system, has several advantages over evolutionary algorithms on the vertex cover problem. However, the analyses only consider a limited number of graph instances.

The purpose of this project is to further investigate evolutionary algorithms and artificial immune systems for the vertex cover problem. The selection of an appropriate benchmark set as well as the comparison of different algorithms and parametrisations on these benchmarks are among the essential tasks. The development of a novel algorithm is optional.

Development and Evaluation of Nature-Inspired Algorithms

At the 10th DIMAC implementation challenge algorithms for Graph Partitioning and Graph Clustering were investigated on a large set of novel benchmark instances. The 9th DIMAC implementation challenge dealt with different versions of shortest paths problems, including dynamic variants, in the very same manner.

The purpose of this project is to develop and evaluate a nature-inspired algorithm for one of these problems. This includes finding an appropriate representation as well as suitable algorithm components.

Computational Intelligence in Games

The goal of this project is to develop an intelligent agent for some game - using an existing framework like:
Christine Zarges