Example PhD

Computer-Assisted Exploration and Model Building for Multi-Variate, High-Dimensional Data Sets

Supervisor: Dr F.C. Langbein

Keywords: Machine learning, pattern recognition, computational modelling, geometry

Measuring and simulating physical and biological systems can easily create huge amounts of high-dimensional, multi-variate data, hard to interpret and understand by humans. Perceptually effective rendering techniques may be used to harness human perception to increase the quantity and clarity of the information conveyed by visualising the data. Although this approach is effective, it still requires that humans interactively explore a potentially large data set, and for huge data sets it is often hard for humans to identify any useful patterns in the data reliably. Physical and biological systems can often be described by partial differential equations and detecting continuous as well as discrete symmetries and similar patterns is a first step to formulate hypotheses explaining the data and derive a computational model from it. However, for large data systems fully automated computer analysis in this fashion is often also hard due to the rather large set of possible alternative explanations. The aim of this project is to combine machine learning techniques to automatically identify interesting features and hypotheses explaining the data with the knowledge and intuition of users exploring the data set to guide its exploration and model building. Of particular interest are systems governed by certain partial differential equations related to nano-technology and biochemical reaction networks.

Key Skills/Background: requires strong mathematical and computing skills

Contact: Dr F.C. Langbein to discuss this research topic.