Learning Sparse Classifiers with Applications to ... - CiteSeerX

7 downloads 52 Views 81KB Size Report
Colin Campbell and Yiming Ying. Department of Engineering ... Work with Colin Cooper (Institute of Cancer Research), Mark Girolami (University of Glasgow),.
Learning Sparse Classifiers with Applications to Biomedical Research Colin Campbell and Yiming Ying Department of Engineering Mathematics University of Bristol [email protected],[email protected]

Abstract In past research we have used unsupervised and semi-supervised methods to resolve cancer expression array datasets into tentative subtypes. These studies have indicated dysregulated genes within these putative subtypes and eligible targets have been investigated using siRNA (small interferring RNA) techniques. Some tentative therapeutic tragets have been found using this approach. A subsequent objective is to fully characterise a gene or protein target of interest. This involves two subsequent objectives. Firstly, investigating functional connections between a given target gene and other genes (network inference) and investigating the functional roles of a given gene (gene and protein function prediction). Both these subtasks involve multi-kernel learning since many types of data are relevant to prediction. We present recent work on multiple kernel learning including novel multi-class and multi-label learning. With these algorithms we attempt to achieve some sparsity among kernel coefficients to indicate which data types are most informative for the given supervised learning task. We additionally attempt to pursue feature weighting to create a hypothesis weighted by the most informative features within each data space. Work with Colin Cooper (Institute of Cancer Research), Mark Girolami (University of Glasgow), Theo Damoulas (University of Glasgow) and others References [1] Colin S Cooper, Colin Campbell and Sameer Jhavar. Mechanisms of Disease: biomarkers and molecular targets from microarray gene expression studies in prostate cancer. Nature (Clinical Practice Urology) Volume 4 (2007) 677-687. [2] Theodoros Damoulas, Yiming Ying, Mark Girolami and Colin Campbell. Inferring Sparse Kernel Combinations and Relevance Vectors: An application to subcellular localization of proteins. Proceedings of the Seventh International Conference on Machine Learning and Applications (ICMLA’08), San Diego, California. [3] Luke Carrivick, Simon Rogers, Jeremy Clark, Colin Campbell, Mark Girolami and Colin Cooper. Identification of Prognostic Signatures in Breast Cancer Microarray Data using Bayesian Techniques. Journal of the Royal Society: Interface Vol. 3 (2006) pages 367-381. [4] Simon Rogers, Mark Girolami, Colin Campbell and Rainer Breitling. The Latent Process Decomposition of cDNA Microarray Datasets. IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2005, Vol. 2, pages 143-156. [5] Y. Li, C. Campbell and M. Tipping. Bayesian automatic relevance determination algorithms for classifying gene expression data. Bioinformatics 2002 18: 1332-1339.

1

Suggest Documents