From Biology to Algorithms

This thesis describes a path from a model of a biological system to a biologically-inspired algorithm. The thesis commences with a discussion of the principled design of biologically-inspired algorithms. It is argued that modelling a biological system can be tremendously helpful in eventual algorithm construction. A proposal is made that it is possible to reduce modelling biases by modelling the biological system without any regard to algorithm development, that is, with only concern of understanding the biological mechanisms. As a consequence the thesis investigates a detailed model of T cell signalling process. The model is subjected to stochastic analysis which results in a hypothesis for T cell activation. This hypothesis is abstracted to form a simplified model which retains key mechanisms. The abstracted model is shown to have connections to Kernel Density Estimation, through developing these connections the Receptor Density Algorithm is developed. By design, the algorithm has application in tracking probability distributions. Finally, the thesis demonstrates the algorithm on a related but different problem of detecting anomalies in spectrometer data