David Duvenaud

Canada Research Chair in Generative Modelling

Tier 2 - 2017-11-01
University of Toronto
Natural Sciences and Engineering Research Council


Research involves

Using a combination of deep learning and statistics to build and fit meaningful computer models from raw data.

Research relevance

This research will create new methods of automating tasks to speed up the discovery and production of sustainable energy products, medicines and robotic controllers.

Automating Science

As the use of artificial intelligence in our every day lives becomes a reality, we gain the ability to do new things based on machine learning—algorithms that enable computers to learn and make predictions. From email and social media platforms that filter messages to speech recognition software, machine learning automatically finds patterns in large amounts of data.

Dr. David Duvenaud, Canada Research Chair in Generative Modelling, designs general learning algorithms for computers so they can fit flexible models directly from unfiltered data and build abstract representations on their own—creating powerful and flexible tools for making predictions or decisions.

Specifically, Duvenaud and his research team are developing ways to automate scientific modelling and theory, and also trying to automate parts of the process of designing solutions to engineering problems.

Working with industrial partners, he has already created programs that can screen and propose new chemical compounds that can benefit the sustainable energy sector, such as in organic LEDs or photovoltaics. He has also developed programs that can automatically classify animal behavioural data directly from video, greatly reducing hands-on notation time for researchers.

By pushing the boundaries of computer science, Duvenaud is contributing to our understanding of the human mind. How we combine information, weigh evidence and find patterns to predict outcomes is at the heart of artificial intelligence research—offering insights into how we learn language, make sense of data and understand the world around us.