Best Student Paper: AI 2020May 19, 2020
We would like to congratulate Xinyu Yun and Tanner Bohn for being selected to receive the best student paper award by Canadian AI 2020 organizers titled “A Deeper Look at Bongard Problems”. Xinyu and Tanner are supervised by Professor Ling.
The Bongard problem (https://en.wikipedia.org/wiki/Bongard_problem) is easy for most humans but very hard for deep learning. For example, given 6 tiles on the left belonging to one class, 6 tiles on the right belonging to another class, which side the test tiles belong to? You can (very likely) identify the underlying patterns with only 12 such training data, but standard deep neural networks would need millions. This work is a bold step towards solving this type of abstract visual learning problems with deep neural networks with a small number of training data - an important machine learning research called few-shot learning.