ECE Seminar: Tensor Analysis and Learning for Medical Data by Dr Haiping Lu: HWLL313, 2-3PM

Tensor Analysis and Learning for Medical Data

Abstract

Big data need machine learning. With so many success stories of deep learning, what are the remaining challenges? This talk will focus on tensor-based machine learning methods for extracting useful information from multidimensional data, such as (f)MRI in brain/medical imaging. The key challenge is to deal with very high dimensionality and multidimensional structures with a very small number of task-specific examples in prediction and interpretation. We will introduce multilinear subspace learning (MSL) for dimensionality reduction, where we learn compact features directly from tensor representations of multidimensional data. Then, we will examine a popular MSL method, multilinear principal component analysis (MPCA). Next, we will present two whole-brain fMRI analysis methods for prediction and interpretation. Finally, we will talk broadly about our current research topics and future directions.

Biography

Haiping Lu is a Lecturer in Machine Learning at the University of Sheffield. He received the BEng and MEng degrees from Nanyang Technological University, Singapore, in 2001 and 2004, and the PhD degree in electrical and computer engineering from the University of Toronto in 2008.

His current research focuses on machine learning for big data in medicine and healthcare, particularly tensor analysis and learning for medical imaging data. He is the leading author of the book “Multilinear Subspace Learning: Dimensionality Reduction of Multidimensional Data” (CRC Press, 2013). He is the recipient of the 2013 IEEE CIS Outstanding PhD Dissertation Award, an awardee of the 2014–2015 Early Career Award by Research Grants Council of Hong Kong, and an awardee of the AAAI-18 Outstanding Program Committee Member Award.

Presented by

Key learning outcomes