NSERC Industrial Research Chair in Machine Learning

Industrial Research Chairs program

Senior Chairholder since 2018

Machine learning has emerged as an essential tool in the quest to build more intelligent computer systems. Machine-learning systems are not programmed to solve a specific problem; rather, they develop their own programming based on examples of how they should behave. Over the last decade, considerable progress has been made, led by advances in deep neural networks. Deep learning techniques are achieving state-of-the-art performance on difficult tasks such as object recognition, speech recognition, and drug discovery. Systems that use deep learning have now made their way into consumer products, such as speech-based intelligent agents on mobile devices.

The proposed research program will focus on bridging two types of methods:

1. those that are capable of scaling up to very large datasets (e.g. deep networks)
2. those that incorporate domain knowledge.

A second focus will be user-controlled learning. Deep neural networks are notoriously difficult to train, and, once they have been trained, it is difficult to understand how they arrive at their responses. As these systems become more prevalent in real-world applications, it is essential to allow users to exert more control over the learning system. Advances stemming from the proposed program are likely to have far-reaching impacts in various areas, including improved scene understanding (combining information from images, video and text) and better electronic health records. A different sort of impact concerns fairness in automated decision systems. The proposed program will develop technical solutions to allow policy-makers and legal experts to exert user control over complex everyday decisions.

Email: zemel@cs.toronto.edu 

Website: 
http://www.cs.toronto.edu/~zemel