Robust learning in computer vision–by design, and by optimization

Speaker: Pradeep Shenoy

Abstract

This talk will cover two recent sets of results – one on using inductive biases (i.e., design) in the development of robust, efficient computer vision systems, and the other on learned reweightings of training loss (i.e., optimization) for achieving a range of robustness properties in classifiers. In the first half, we look at how simple, well-chosen inductive biases can significantly impact accuracy of vision models, with case studies in object categorization and multi-object multi-part segmentation. The second half will cover the paradigm of learned reweighting using meta-learning for robust classification. We will cover recent developments in this framework for tackling learning under concept drift, selective classification, and domain generalization.

Bio

Pradeep Shenoy leads the Cognitive modeling & Machine Learning team at Google Research India, which aims to build expressive, robust ML systems, drawing functional and algorithmic inspiration from human cognition. Pradeep also works on modeling human behavior & cognition, with applications in personalization and human-AI interfaces. Recent work has focused on robust learning via instance reweighting, and its application to a range of problem settings in applied ML. Pradeep has a Ph.D. in Computer Science from the University of Washington & post-doctoral research experience at UC San Diego, where he worked in neuro-engineering, computational neuroscience & cognitive science. He has previously led machine learning teams at Microsoft Bing, developing and supporting large-scale production models that predict user behavior (clicks, conversions, audience segmentation, etc.) in sponsored search.