Sequential pattern induction for bioinformatics

点击翻译按钮获取中文摘要

The proposed research applies a number of machine learning methods to problems of bioinformatics.Emphasis is on learning algorithms that learn to classify, generate and transmute sequences of symbols basedon example sequences. Over the past 10 years a number of new approaches to inducing sequences have beenproposed by myself and other researchers in the field. These have been sparsely applied to selectedbioinformatics problems and analysed.The objectives of this research are to discover which techniques work best on which problems, and how thesemethods can be adapted to the unique constraints arrising from practical problems on real biological datasets.This will allow scientists working in bioinformatics to judiciously apply these techniques to problems ofinterest as well as provide some insight into alternative approaches that might not otherwise be considered. Itwill effect an easier transfer of knowledge from research in machine learning to practical bioinformaticsproblems.The research will focus on (but not be limited to) the following techniques: artificial neural networks,automata models, formal grammars and hidden Markov models.The research is a follow-up to-- and extension of-- previous work in this area at a much enhanced scale. Thisincrease is due in part to the recent accredidation of my department to grant PhD degrees and three PhDstudents under my supervision.