Skip to content. Skip to navigation
CIM Menus

Department of Electrical and Computer Engineering Research Seminar

Dimensionality Reduction, Classification and Applications in Healthcare

Narges Armanfard
University of Toronto

April 12, 2018 at  11:00 AM
McConnell Engineering Room 603


The Dimensionality reduction is a very important component in data classification applications. It is an antidote to what Bellman referred to as the “curse of dimensionality”. It is well known that the performance of typical classifiers notably drops when the number of available observations is not adequate in comparison to the number of candidate features.

Feature selection approaches perform dimensionality reduction by selecting a subset of relevant features (from the available set of candidate features) that leads to an “optimal” characterization of different classes. Conventional feature selection algorithms select a single common feature set for characterizing all regions of the sample space. In fact, these methods assume that a single feature subset can optimally characterize sample space variations.

In this talk, I will present an alternative view to the traditional concept of a common feature set. I will discuss the novel concept of localized feature selection whereby each region of the sample space is associated with its own distinct optimized feature set, which may vary both in membership and size across the sample space. This allows the feature set to optimally adapt to local variations in the sample space. A localized classifier for measuring the similarities of a query datum to each of the respective classes will also be presented.

Furthermore, I will discuss relevant applications including automatic and continuous Mismatch Negativity (MMN) detection for coma outcome prediction and zero-effort technologies for unobtrusive home-based vital parameters estimation.


Narges Armanfard is currently a postdoctoral fellow at Intelligent Assistive and Technology Lab at the University of Toronto under supervision of Prof. Alex Mihailidis. She received her PhD degree in Electrical and Computer Engineering from McMaster University in 2016. Her research interest and experience are in the areas of machine learning and its applications in healthcare. She is the author of 30 peer-reviewed articles (including TPAMI, TNNLS, TCYB and Pattern Recognition), and two patents.