Dr Naeemullah Khan, King Abdullah University of Science and Technology
Abstract
Machine learning algorithms are showing remarkable success across multiple domains. The key ingredients for the success of machine learning methods are large data, and large networks. Learning large model in large dataspaces, inevitably create lack of robustness. In my talk, I will present an overview of the generalization issues of the current state-of-the-art, techniques used to introduce robustness and some of our work on robustness to deformation for images.
