Machine Learning, AI, Data Science November 20, 2020

Event Video

Panel Discussion Video


Chris J. Maddison

U of T, DeepMind

Chris J. Maddison is an Assistant Professor of Computer Science and Statistics at the University of Toronto. He is the acting CIFAR AI Chair at the Vector Institute and a Research Scientist at DeepMind.

“My goal is to understand and improve the algorithms that agents can use to learn from data and reason about their experience. Learning can be formalized in the language of statistics. Because statistics usually involves solving difficult problems, like probabilistic inference or optimization, most learning systems rely on algorithms for these problems. Of these, I am particularly interested in algorithms for (approximate) Bayesian inference, Monte Carlo estimation, and continuous and discrete optimization. Although these problems seem distinct, they have a lot of shared structure. My work often touches on this theme, like when we showed how to simulate from a probability distribution by optimizing a random function. I am also interested in learning with structured data. Together with colleagues, we built the first artificial agent that plays the board game Go at a superhuman level, developed structured models of human-written source code, improved the training of latent variable models of time series, and designed relaxed gradient estimators for models with structured latent variables.”

Gradient Estimation with Stochastic Softmax Tricks

Gradient estimation is an important problem in modern machine learning frameworks that rely heavily on gradient-based optimization. For gradient estimation in the presence of discrete random variables, the Gumbel-based relaxed gradient estimators are easy to implement and low variance, but the goal of scaling them comprehensively to large combinatorial distributions is still outstanding. Working within the perturbation model framework, we introduce stochastic softmax tricks, which generalize the Gumbel-Softmax trick to combinatorial spaces. Our framework is a unified perspective on existing relaxed estimators for perturbation models, and it contains many novel relaxations. We design structured relaxations for subset selection, spanning trees, arborescences, and others. We consider an application to helping make machine learning models more explainable.

Mike Domaratzki

University of Manitoba

Mike Domaratzki is an Associate Professor of Computer Science at the University of Manitoba. He has a variety of research interests in bioinformatics, data science and machine learning. His current research uses deep learning to predict crop performance and drive crop improvement.

Predicting with Rare Data 

How do we predict events that happen rarely? In many datasets, the outcome we are interested in is rare. Examples include disease in humans, complications in surgery or fraud in financial records.   In this talk, we will discuss how balancing a dataset is important for machine learning. We will look at how creating synthetic data instances can improve prediction for machine learning tools by looking at an example from customer retention. 

Dr. Anastasia (Stasi) Baran

COO, nQube Data Science

Dr. Stasi Baran is a Co-founder and COO of nQube Data Science. She received her PhD in Electrical and Computer Engineering from the University of Manitoba in 2016, where she specialized in applications of non-linear optimization methods. She holds a B.Sc. in physics and an M.Sc. in astrophysics. Stasi uses her multi-disciplinary background to find and develop suitable applications for nQube’s advanced artificial intelligence technology in the casino industry. Her combined interests in large-scale data modeling problems and the gaming industry have helped to develop nQube’s AI-based slot floor optimization and player segmentation solutions.

AI in the casino gaming industry – applications in an unexpected, data rich field

nQube Data Science specializes in artificial intelligence solutions in the casino gaming industry. Our flagship Reel AI solution uses AI-driven evolutionary computing techniques to optimize slot floors by telling casinos which slot machines to buy, which ones to retire, and where to put them on the slot floor. Due to regulatory obligations, casinos have been collecting large quantities of extremely granular, transactional level data for decades. Casino-based data scientists and academics at the University of Nevada Las Vegas have been studying this unique “casino laboratory” for years. The data rich nature of this industry has now expanded to other academic institutions and innovative companies looking to modernize the casino industry, increase revenue for casino operators and tribal economies, and protect those who are prone to unhealthy gambling behaviour. This presentation will provide an overview of the AI and data science opportunities that exist in this field, as well as the challenges around breaking into a closed, highly regulated, and secretive industry. nQube’s innovative applications in the space will be demonstrated, and casino gaming science as it advances as an academic field will be discussed.

Matt Schaubroeck

CEO, ioAirFlow

Matt Schaubroeck is CEO and co-founder of ioAirFlow. He completed a Master’s in Business Administration at the University of Manitoba, with a focus on entrepreneurship. He has been working on ioAirFlow since 2016, which started as a class project. He moved to the company full-time in early 2020.

ioAirFlow is providing a digital audit platform that helps ESCOs provide faster, cheaper, and more accurate audits for commercial buildings. Using AI software and secure IoT sensors, ioAirFlow helps diagnose a building’s efficiency problems and recommend actionable solutions to resolve any existing problems.

So hot right now:
Using big data analytics and IoT sensors to identify building performance, efficiency problems, and help mitigate climate change

Almost every building in the world is less efficient than it should be. In these cases, their operations and utility costs are rising, tenants are spending time in an uncomfortable or unsafe environment, and greenhouse gas emissions are being unnecessarily generated. There are no known shortages of efficiency solutions for buildings to improve their performance. However, financial and informational barriers often prevent building owners from identifying the problems that exist in their properties. What remains is a global commercial building efficiency gap. How can we solve this massive global problem? Join ioAirFlow to learn about how cutting-edge hardware and intelligent analysis software will revolutionize energy audits and provide another tool to mitigate climate change.

Dr. Lilian Wong

Senior Applied Scientist, Amazon Web Services

Dr Lilian Wong is Senior Applied Scientist with Amazon Web Services (AWS). Her recent work relates to the applications of deep learning to forecasting and anomaly detection, especially in the renewable energy domain. In her previous role in AWS, she has built the forecasting model that powers Predictive Scaling for EC2, which uses deep learning to help customers scale their computing fleets according to their workloads.

Dr. Wong holds a PhD in mathematics from the California Institute of Technology, specializing in analysis and mathematical physics. Like many theorists, she did not have any coding or machine learning experience until she decided to apply for a job in the industry.

“Life Outside of the Bubble”

 Many mathematicians have not really pondered the possibilities outside of the academia until they had to leave it. I will talk about some of the career options for mathematicians (and other theorists) and how to land a job in the industry

Luz Angélica Caudillo Mata

R&D Scientist, MDA

Luz Angélica Caudillo-Mata is a Computational Mathematician. Luz Angélica’s professional journey has been shaped by the question: how can one develop and apply mathematics to solve real-world problems? In pursuit of an answer, she gained knowledge in the fields of Mathematics, Computer Science, and Geophysics at the bachelor’s, master’s, doctoral, and postdoctoral levels. She has also conducted applied multidisciplinary research at top international institutions, such as Lawrence Berkeley and Livermore National Labs in the USA; The University of British Columbia in Canada; the Polytechnique University of Valencia, and the Complutense University of Madrid in Spain; and the Mathematics Research Center in Mexico.

Luz Angélica’s expertise is in numerical analysis and scientific computing. She specializes in the design and implementation of computational methods and algorithms for partial differential equations (PDE’s), PDE-constrained optimization, computational inverse problems, and machine learning. Some of the applications she has worked on are geophysical modeling and prospectivity problems for natural resource (minerals, oil, groundwater) exploration programs, and mechanical structural analysis and optimization for affordable roof construction. As a researcher, Luz Angélica has published her work at top computational mathematics peer-reviewed journals; collaborated in developing mathematical technology and services for industrial applications; presented her research outcomes at numerous high-profile international conferences in computational mathematics, scientific computing, geophysics, and engineering; and co-organized 50+ technical events. She has also co-founded 3 institutional programs to support the diversity of connections among STEM areas, and the mathematical community as a whole.

Currently, Luz Angélica works as an R&D Scientist at MDA Corporation, the Canadian space-tech company behind the iconic Canadarm, where she develops Machine Learning algorithms using satellite imagery and computer vision techniques to develop the next generation of Earth monitoring systems.

Making a Career Transition into Industry:

Stories & Tips from Math, CS, and Stats Alumni

In this panel, alumni of graduate programs in mathematics, computer science, and statistics share how they successfully transitioned from academia into industrial positions in the areas of machine learning, AI, and data science. After learning from the story of each of our four panelists, we’ll have an extensive Q&A session.


Dr. Roger Donaldson

Roger is a principal scientist working at Motorola Solutions, Inc.  Roger has always enjoyed mathematics, but only seriously considered it as a career after meeting Brian Wetton of the UBC Mathematics Dept who at the time was doing an industrial collaboration with Ballard Power, a Burnaby fuel cell company, an early Mitacs project.  Roger ultimately completed a Ph.D. at Caltech, worked at Google, and then returned to Vancouver to work as an industrial mathematics consultant.

Roger’s current role at Motorola Solutions has him leading a team of data scientists and engineers publishing machine learning models that combine video and language to the end of helping first responders get the information they need in moments that matter.

Dr. Andrés Muñoz Medina

Andrés is a mexican researcher working at Google. Andrés love for mathematics started early on while participating in the math olympiads as a high-school student. He then received a bachelor’s degree in Mathematics from UNAM in Mexico City. After his bachelor, Andrés received a Ph. D. from the Courant Institute of Mathematical Sciences at New York University. While getting his degree, Andrés interned at Google. 

Andrés interests are in theoretical machine learning and statistics. He has done a lot of research on algorithmic game theory and auction mechanism design. More recently, Andrés has begun working on differential privacy and making sure that user data is protected at Google. 


%d bloggers like this: