Introducing Applied Deep Learning at UofT

A New Course for Graduate Students at The University of Toronto

The University of Toronto is the birthplace of deep learning, and also one of the world’s foremost centres for research across the board.

It’s official!

Today we’re excited to announce a new course for graduate students at the University of Toronto, facilitated by Ragavan Thurairatnam, our Chief of Machine Learning, and Jodie Zhu, one of our Machine Learning Engineers.

Ragavan and Jodie were recently appointed as Adjunct Lecturers at the University’s Dalla Lana School of Public Health, where they will teach a course on Applied Deep Learning to graduate students studying at the school (as well as those from other faculties). In addition, the course will also feature a number of guest lecturers from our Machine Learning Engineering team throughout the semester.

Ragavan is a pioneer when it comes to applying deep learning in industry.

We think this course will fill an important gap in the current academic offerings for students advancing their knowledge of machine learning. At UofT and other institutions today, the majority of courses focus almost exclusively on the theoretical underpinnings of the field. With this course, Ragavan and Jodie are setting out to change that.

Both Ragavan and Jodie are passionate about making it easier to acquire the skills needed for real-world deep learning. And they’re each uniquely qualified for this task as well.

Ragavan, for example, has been developing deep learning applications for industry since 2012, the year that AlexNet ushered in what many people call deep learning’s “Cambrian Explosion.” Since then, Ragavan has helped some of the world’s biggest companies save millions of dollars in revenue and build the capabilities required for scaling AI throughout their organizations. Earlier this year, he also shared the stage with AI’s godfather, Geoffrey Hinton, at a panel organized by UofT students focussed on distinguishing between AI hype and reality.

Just years into her career, Jodie has already made an enormous impact applying data science and deep learning to multiple industries, ranging from telecommunications to video game development. With a Masters of Science in Biostatistics from UofT, she has also extensively explored deep learning’s applications for public health. Before joining Dessa, Jodie worked at the Princess Margaret Cancer Centre, where she co-authored and published multiple papers on advancing breast cancer treatment using data science.

Drawing from their extensive experiences deploying real-world AI, their course will help students develop the skills needed to put deep learning into production.

Because it will be taught at the Dalla Lana School for Public Health, the course will also zoom in on the enormous potential AI techniques like deep learning have for improving healthcare at a broad societal level.

The first session of this course will kick off this summer, running from June 17 to August 16. Current graduate students can learn more about how to register for the course by checking out the 2019 Summer Timetable at the Dalla Lana School’s website.

To learn more about the course, our Communications Lead Alyssa sat down with Jodie, asking her a few questions about why the course is such a unique opportunity for students.

Jodie presenting on “Neural Ordinary Differential Equations,” one of the field’s biggest papers last year, at the Toronto Deep Learning Series

AK: Can you talk a bit more about the way the course at UofT differs from other courses that are currently available on machine learning?

JZ: There’s this huge gap that exists right now between the level of experience graduate students have with machine learning and the experience that’s required if you want to actually make a difference with it in industry.

So far, we think that’s because there are skills that to date could have only been acquired by working on industry problems, versus just on research. We have noticed this when reviewing applications for machine learning positions from recent graduates. They get the theory, but they just haven’t had the experience working with messier types of data, as an example. They don’t know what the machine learning workflow is like from end to end.

After finishing our course, students will be ready to tackle the kinds of machine learning problems they’ll encounter in real-world settings.

AK: Can you us a bit of insight into what the curriculum will look like?

The course will be half lectures, half workshops. We’ll teach students how to translate a real-world problem into the language of machine learning. And talk about the unique kinds of infrastructure they’ll encounter when working on practical applications. Students will also learn the basics of putting practical applications into production, and how there is a lot more to real-world machine learning that just writing code.

To make sure all participants have the right foundation, we’ll also devote time in the first few weeks of the course to go through the fundamentals of neural networks and deep learning in a more theoretical way.

The biggest way we’ll be evaluating the course is with a final project, where students will work in small groups to develop a real-world application for deep learning related to public health.

AK: It’s amazing that the course is being offered at the Dalla Lana School for Public Health, since there are so many applications for deep learning in healthcare. What are your thoughts on the applications in this field?

One of the reasons Ragavan and I chose to approach this faculty about the course is because of this very reason. We both see how much of an impact deep learning could have on improving healthcare around the world.

When we talk about using AI for health, I think a lot of people immediately jump to thinking about glamorous applications like using image recognition to diagnose rare diseases, or for drug development. These applications obviously have amazing potential. But then there are also these less glamorous applications which are more behind the scenes, and I see tremendous value there too. Things like reducing waiting times in hospitals.

For me, these kinds of practical applications are actually some of the things that I’m most fascinated by. I love being able to see the impact of my work and how it can improve existing processes. In our work with enterprises, we’ve already seen how making a good prediction with machine learning can lead to millions of dollars in savings in a year. Now we want to explore how we can work with public health to make an impact there, and working with these students is definitely one way to do that.


Wish you could learn more about Dessa, applied deep learning, or more future sessions of this course at UofT? You’re in luck: we publish a monthly newsletter that covers all of the things you need to know, not only about applied AI but also the latest developments at Dessa. Check out the latest issue and subscribe here. Already convinced? Find a direct link to subscribe to the newsletter here.