Sandra Zilles


Canada Research Chair in Computational Learning Theory

Tier 2 - 2017-11-01
Renewed: 2017-04-01
University of Regina
Natural Sciences and Engineering Research Council

306-337-3250
zilles@cs.uregina.ca

Research involves


Designing and analyzing new models and algorithms for interactive machine learning.

Research relevance


This research will lead to the development of intelligent data analysis methods for medical science, bioinformatics, e-commerce and other areas.

Helping Machines Learn More Quickly Through Co-Operation


Machinelearning is about designing algorithms that help computers “learn” based ondata. For example, it is used to identify user preferences in web searches,personalize advertising on e-commerce sites or develop robots that can learnfrom each other.

Unfortunately,standard machine learning models don’t benefit from the fact that all thesescenarios have one thing in common: interaction with a co-operative partner.

Currentmachine-learning models assume that machines learn from random data. But Dr.Sandra Zilles, Canada Research Chair in Computational Learning Theory,studies models in which machines learn from well-chosen and specific data, asthough they were interacting with a “teacher.” She compares her work toclassroom learning, where students learn using materials carefully selectedby the teacher.

Theadvantage to interactive machine learning is that it requires less data thancurrent models. Through research, Zilles will make intelligent machinesexploit the quality of chosen data rather than process large quantities ofpotentially expensive data.

The modelsand algorithmic techniques that Zilles and her research team provide maycompletely change the way machine learning is deployed. It may also provideefficient solutions to complex problems in artificial intelligence at a lowercost and with less data than is currently possible.