Skip to content. Skip to navigation
McGill Home SOCS Home
Personal tools
You are here: Home Announcements and Events Colloquium Archive

Colloquium Home
Fall 2013 Schedule
Winter 2014 Schedule
How to Suggest a Speaker


( Winter 2014 )
Speaker and Abstract
2014/01/31 Speaker: Cathy Laporte
Affiliation: École de Technologie Supérieure
Area: Medical Image Analysis
Title: Ultrasound image analysis for applications in speech science
Abstract: Ultrasound (US) imaging is an effective and non invasive way of observing the tongue motions involved in normal and pathological speech, and the results of US studies are of interest to the study of the mechanisms underlying speech as well as the development of new strategies in speech therapy. (Semi-)automatically segmenting the tongue contour as it evolves in ultrasound video sequences is an image analysis task of particular interest since it allows the systematic study of tongue shape and motion. Challenges include maintaining high quality tongue tracking over time, as well as monitoring segmentation quality and developing meaningful ways to characterize tongue shape and motion. Our preliminary results touch upon each of these challenges and show how new adaptations of techniques previously rooted in the computer vision literature can be combined to provide useful tools to the speech scientist. Biography of Speaker:

Catherine Laporte obtained her Ph.D. from McGill University in 2010 and is now a professor at the department of electrical engineering at École de technologie supérieure (ÉTS). She teaches courses in medical imaging, biomedical instrumentation and algorithms. Her research group focuses on the development of new methods for the analysis of ultrasound images to address problems such as 3-D reconstruction, motion and deformation tracking, segmentation and registration, with applications in orthopaedics and speech science.

2014/02/14 Speaker: Bruce Reed
Affiliation: McGill University
Title: How I Learned to do Mathematics
Abstract: I was exceedingly fortunate to have Vašek Chvátal as a professor when I was an undergraduate at McGill and as my supervisor as a graduate student. His instruction as to how to carry out mathematical research has stood me in good stead, ever since. I will discuss what I learnt from him and how I applied it throughout my career. Biography of Speaker:

Professor Reed received his degrees in Mathematics and in Computer Science at McGill University. Following postdoctoral fellowships and faculty positions in Europe, Canada and the USA, he joined the faculty of McGill University in 2001. He currently holds the Canada Research Chair in Graph Theory and was elected as a Fellow of the Royal Society of Canada in 2009. Professor Bruce Reed has been awarded the 2013 CRM-Fields-PIMS Prize. The Prize recognizes exceptional achievement in the area of mathematical sciences and is considered one of Canada’s top honours in mathematics.

2014/02/28 Speaker: Lukasz Golab
Affiliation: University of Waterloo
Title: Optimizing data placement for distributed computation
Abstract: I will discuss the following problem: given a set of data items, a set of tasks that reference these data items, and a set of servers with finite storage and processing capacities, allocate the data items to the servers in a way that minimizes the amount of data that needs to be transferred among servers during task execution. This problem arises in many practical scenarios including cloud databases. I will show that this problem can be reduced to the well-studied graph partitioning problem, which is NP-hard, but for which efficient approximation algorithms exist. I will also discuss how to handle load balancing and data replication. This is joint work with Marios Hadjieleftheriou (AT&T), Howard Karloff (Yahoo!) and Barna Saha (AT&T). Biography of Speaker:

Lukasz Golab is a faculty member in the Management Sciences department at the University of Waterloo. Previously he was a Senior Member of Research Staff at AT&T Labs. He obtained a BSc from the University of Toronto in 2001 and a PhD from the University of Waterloo in 2006. His research interests are in database systems, data mining and energy data management.

2014/03/14 Speaker: Aaron Fenster
Affiliation: Robarts Research Institute, The University of Western Ontario
Title: Use of 3D Ultrasound Imaging in image-guided interventions
Abstract: The last two decades have witnessed unprecedented developments of new image-guided interventional imaging systems making use of 3D visualization. These new technologies provide the clinician with guidance and verification information about the interventional procedure. Although 2D ultrasound (US) has been used for image-guidance, this approach limits our ability to guide therapy, because multiple 2D images must be integrated mentally, resulting in an inefficient procedure, which often leads to variability. Investigators have addressed these limitations by developing 3D US techniques. In this paper we describe our developments of 3D US imaging instrumentation and techniques for use in image-guided interventions. In our approach the conventional US transducer is scanned mechanically, and the 2D US images are digitized and reconstructed in real-time into a 3D image, which can be viewed and manipulated interactively. Examples will be given for use in 3D US-guided prostate biopsy and brachytherapy as well as 3D US-guided focal liver ablation. Supported in part by grants from CIHR, ORF and OICR. Biography of Speaker:

Dr. Fenster received his PhD degree in 1976 from the Department of Medical Biophysics of the University of Toronto. In 1987 he moved to London and became a Scientist and founding Director of the Imaging Research Laboratories (IRL) at the Robarts Research Institute and Professor at The University of Western Ontario (UWO) in Radiology. Under his leadership, the IRL has grown to a staff of 250 today. In addition, he is the founder and Associate Director of the interdisciplinary graduate Program at UWO in Biomedical Engineering. He is also the Chair of the basic Science Division of the Department of Medical Imaging and the Director for the Biomedical Imaging Research Centre at UWO. This Division combines the strengths in imaging research across London’s Institutions, which combined, makes the London medical imaging research community one of the largest in North America with over 350 staff and students and $100M in research equipment. In 2010 he became the Centre Director of the newly formed Centre for Imaging Technology Commercialization – a federally funded Centre of Excellence in Commercialization and Research. Currently, he holds a Canada Research Chair-Tier 1 in Biomedical Engineering. He is the first recipient of the Premier’s (Ontario) Discovery Award for Innovation and Leadership (2007), the Hellmuth Prize for Achievement in Research at the UWO (2008), and the Canadian Organization of Medical Physicists (COMP) Gold Medal Award (2010). In 2011 he was inducted into the Canadian Academy of Health Sciences. Fenster’s research has resulted in 270 peer-reviewed publications, 39 patents and the formation of four companies. His patents have been licensed to 13 companies, which have commercialized them for world-wide distribution.

2014/03/21 Speaker: Emma Brunskill
Affiliation: Carnegie Mellon University
Title: Sequential Transfer in Multi-armed Bandit with Finite Set of Models
Abstract: Learning from prior tasks and transferring that experience to improve future performance is critical for building lifelong learning agents. Although results in supervised and reinforcement learning show that transfer may significantly improve the learning performance, most of the literature on transfer is focused on batch learning tasks. In this paper we study the problem of sequential transfer in online learning, notably in the multi-armed bandit framework, where the objective is to minimize the total regret over a sequence of tasks by transferring knowledge from prior tasks. We introduce a novel bandit algorithm based on a method-of-moments approach for estimating the possible tasks and derive regret bounds for it. Biography of Speaker:

Emma Brunskill is an assistant professor in the computer science department at Carnegie Mellon University. She is also affiliated with the machine learning department at CMU. She works on interactive machine learning, focusing on applications that involve artificial agents interacting with people, such as intelligent tutoring systems. For her research Emma was selected as a Microsoft Faculty Fellow.

2014/03/28 Speaker: Tiberiu Popa
Affiliation: Concordia University
Title: Kinectricks: Space-time Capture of Dynamic Objects using Hybrid Cameras
Abstract: The acquisition of moving and deforming objects has lately received a lot of interest in computer graphics and vision due to its many applications in virtual and augmented reality, animation and special effects, 3D video and tele-presence. This is a challenging problem. Some objects such as skin or cloth exhibit very complex dynamic behaviour and fine geometric details difficult to capture. Another critical aspect in acquiring dynamic objects is maintaining temporal coherence between frames. Also due to occlusions the captured objects will often have missing geometry that needs to be reconstructed. In this talk I will present some of my work in space-time capture with several applications in modeling, deformation, novel view synthesis, telepresence and teleconferencing. At the core of most of our hardware setups there are one or more Kinect devices. This makes our setup inexpensive, but also raises a new set of challenges related to the low quality of the Kinect data. Biography of Speaker:

Professor Tiberiu Popa joined the department of Computer Science and Software Engineering at Concordia University in August 2013. He was previously a Senior Researcher at ETH Zurich in the computer graphics lab and he completed my PhD at University of British Columbia. He has been doing research in Computer Graphics for over ten years and has supervised and mentored several Bachelor, Masters and PhD students. His research portfolio is diverse and includes geometric modeling, animation, 2D to 3D reconstruction, among other topics. One important focus of his research is spatio-temporal geometry acquisition (or 4D geometry acquisition) with vast applications in many disciplines ranging from games, engineering, e-commerce to medical.

2014/04/04 Speaker: Hanumant Singh
Affiliation: Woods Hole Oceanographic Institution
Title: Bipolar Robotics: From the Arctic to the Antarctic with a stop at a few places in between
Abstract: In this talk we look at the role of Autonomous Underwater Vehicles (AUVs) and their use for a variety of applications in challenging environments in the Arctic and the Antarctic. In particular we look at the constraints underwater mapping places on robotics in dynamic, ice covered waters. This talk also examines the role of AUVs for Fisheries and segues into some of the underwater imaging issues associated with optical imaging of the seafloor in the context of segmentation, classification and machine learning. Biography of Speaker:

Hanumant Singh completed his Ph.D. in the MIT WHOI Joint Program in 1995 and has since been on the staff at WHOI. His research interests are in the area of underwater imaging and robotics and his work has taken him on more than 50 research expeditions in all of the world's oceans covering a variety of topics including Marine Archaeology, Marine Geology, Marine Chemistry, Coral Reef Ecology and Fisheries, and Glaciology and the Study of Sea Ice.

2014/04/11 Speaker: Greg Wilson
Affiliation: Mozilla Foundation
Title: Two Solitudes
Abstract: I have spent much of the last fifteen years trying to build bridges between the two solitudes of computing: academic researchers on the one side, and working developers on the other. These efforts have largely failed, but have done so in interesting ways. This talk will explore why the wide gulf between research and practice persists, and outline a new plan for trying to narrow it based on scurvy, smoking, and statistics. Biography of Speaker:

Greg Wilson is the creator of Software Carpentry, a crash course in computing skills for scientists and engineers. He has worked for 25 years in high-performance computing, data visualization, computer security, and academia, and is the author or editor of several books on computing (including the 2008 Jolt Award winner "Beautiful Code") and two for children. Greg received a Ph.D. in Computer Science from the University of Edinburgh in 1993, and presently runs the Software Carpentry project for the Mozilla Foundation.