Place: Room 621, GP South (Building 78)

Time: Wed 26th July, 10:30 morning Tea. 11:00am seminar

Title "Classifying sequence data with kernels: How to make really good machine learning algorithms really, really bioinformatics friendly"

Speaker: Mikael Boden, ACCS - ITEE

Abstract
Machine learning algorithms based on "kernels" can be highly versatile tools for biological sequence data analysis. Recent developments include the spectrum kernel, substitution kernel, profile kernel and local alignment kernel, all of which aim to endow support-vector machines with biological knowledge to perform accurate classification of diverse sequence data. This talk reviews support-vector machines and kernels tailor-made for biological sequence data, and highlights their strengths and weaknesses for biological sequence classification. Experiences from approaching various protein classification problems, such as protein localisation in bacteria, peroxisomal protein import signals and sub-nuclear localisation, are discussed.