Timezone: »

Visual Concept Learning: Combining Machine Vision and Bayesian Generalization on Concept Hierarchies
Yangqing Jia · Joshua T Abbott · Joseph L Austerweil · Tom Griffiths · Trevor Darrell

Sun Dec 08 02:00 PM -- 06:00 PM (PST) @ Harrah's Special Events Center, 2nd Floor

Learning a visual concept from a small number of positive examples is a significant challenge for machine learning algorithms. Current methods typically fail to find the appropriate level of generalization in a concept hierarchy for a given set of visual examples. Recent work in cognitive science on Bayesian models of generalization addresses this challenge, but prior results assumed that objects were perfectly recognized. We present an algorithm for learning visual concepts directly from images, using probabilistic predictions generated by visual classifiers as the input to a Bayesian generalization model. As no existing challenge data tests this paradigm, we collect and make available a new, large-scale dataset for visual concept learning using the ImageNet hierarchy as the source of possible concepts, with human annotators to provide ground truth labels as to whether a new image is an instance of each concept using a paradigm similar to that used in experiments studying word learning in children. We compare the performance of our system to several baseline algorithms, and show a significant advantage results from combining visual classifiers with the ability to identify an appropriate level of abstraction using Bayesian generalization.

Author Information

Yangqing Jia (Alibaba Group)
Joshua T Abbott (UC Berkeley)
Joseph L Austerweil (University of Wisconsin, Madison)

As a computational cognitive psychologist, my research program explores questions at the intersection of perception and higher-level cognition. I use recent advances in statistics and computer science to formulate ideal learner models to see how they solve these problems and then test the model predictions using traditional behavioral experimentation. Ideal learner models help us understand the knowledge people use to solve problems because such knowledge must be made explicit for the ideal learner model to successfully produce human behavior. This method yields novel machine learning methods and leads to the discovery of new psychological principles.

Tom Griffiths (Princeton)
Trevor Darrell (UC Berkeley)

More from the Same Authors