Deep Learning with Clear Semantics

Speaker:        Dr. Pascal Poupart
                University of Waterloo

Title:          "Deep Learning with Clear Semantics"

Date:           Friday, 18 November 2016

Time:           11:00am - 12 noon

Venue:          Lecture Theater F (near lifts 25/26), HKUST

Abstract:

Deep neural networks have led to important advances in several
applications including breakthroughs in speech recognition and computer
vision.  However, it is not always clear how to interpret the quantities
computed by each node in most neural networks. In this talk, I will
discuss a special type of neural network known as sum-product networks for
which the semantics are clear. More precisely, I will explain how
sum-product networks can be interpreted as probabilistic graphical models
such as Markov networks and Bayesian networks.  I will also explain how to
interpret the quantities computed by each node as probabilities with
respect to certain variables that correspond to features in the data.  The
probabilistic interpretation of sum-product networks also opens the door
to several learning algorithms (e.g., expectation maximization, signomial
programming, Bayesian moment matching, variational inference) that do not
suffer from the gradient vanishing problem and are much more effective
than stochastic gradient descent (a.k.a. backpropagation). Furthermore,
sum-product networks support a wide range of queries where any feature can
serve as input, output or simply be omitted, while most neural networks
can answer only a single type of query based on a predetermined set of
input and output features.  Examples of sum-product networks will be
presented in various application domains.

*******************
Biography:

Pascal Poupart is an Associate Professor in the David R. Cheriton School
of Computer Science at the University of Waterloo, Waterloo (Canada).  He
received the B.Sc. in Mathematics and Computer Science at McGill
University, Montreal (Canada) in 1998, the M.Sc. in Computer Science at
the University of British Columbia, Vancouver (Canada) in 2000 and the
Ph.D. in Computer Science at the University of Toronto, Toronto (Canada)
in 2005.  His research focuses on the development of algorithms for
reasoning under uncertainty and machine learning with application to
health informatics and natural language processing.  He is most well known
for his contributions to the development of approximate scalable
algorithms for partially observable Markov decision processes (POMDPs) and
their applications in assistive technologies, including automated
prompting for people with dementia for the task of handwashing. Other
notable projects that his research team are currently working on include
chatbots for automated personalized conversations and wearable analytics
to assess modifiable health risk factors.  He co-founded Veedata, a
startup that provides analytics services to the insurance industry and the
research market.

Pascal Poupart received a Distinguished Collaborator Award from Huawei
Noah's Ark in 2016, a David R. Cheriton Faculty Award in 2015 and an Early
Researcher Award (competitive honor for top Ontario researchers) by the
Ontario Ministry of Research and Innovation in 2008.  He was also a
co-recipient of the Best Paper Award Runner Up at the 2008 Conference on
Uncertainty in Artificial Intelligence (UAI) and the IAPR Best Paper Award
at the 2007 International Conference on Computer Vision Systems (ICVS). He
also serves on the editorial board of the Journal of Machine Learning
Research (JMLR) (2009 - present) and the Journal of Artificial
Intelligence Research (JAIR) (2008 - 2011). His research collaborators
include Huawei, Google, Intel, Kik Interactive, In the Chat, Slyce,
HockeyTech, the Alzheimer Association, the UW-Schlegel Research Institute
for Aging, Sunnybrook Health Science Centre and the Toronto Rehabilitation
Institute.