The 2020 edition of the Conference on Uncertainty in AI (UAI) was held last week, bringing together academic and industrial researchers and students from a host of fields including statistics, data science, AI, probabilistic reasoning, and decision making.

One of the “big five” conferences on machine learning, UAI focuses on knowledge representation, learning, and reasoning in the presence of uncertainty. A concept inherent to machine learning, uncertainty refers to the imperfect or incomplete information with which machine learning researchers must work.

Originally slated to be held in Toronto, this year’s conference was held virtually, with a number of Vector Faculty Members taking key leadership positions. Roger Grosse was local arrangements chair, Pascal Poupart sat on the senior program committee, and Faculty Member David Duvenaud served as sponsorship chair with the help of Mona Davies and other members of the Vector professional staff.

“Every year UAI attracts some of the most prestigious researchers in the field,” says Duvenaud, noting that there were more than 400 paper submissions. “I myself have published some of my favourite papers at this conference.”

Several of this year’s talks also included Vector Faculty members:

• Jimmy Ba – “Learning Intrinsic Rewards as a Bi-Level Optimization Problem”
• Animesh Garg – “OCEAN: Online Task Inference for Compositional Tasks and Context Adaptation”
• Pascal Poupart – “Batch norm with entropic regularization turns deterministic autoencoders into generative models”

Next year’s edition is currently scheduled to be held in Toronto.