16891

Video URL

https://simons.berkeley.edu/talks/recent-developments-supervised-learning-noise

Recent Developments in Supervised Learning With Noise

(2020). Recent Developments in Supervised Learning With Noise. The Simons Institute for the Theory of Computing. https://simons.berkeley.edu/talks/recent-developments-supervised-learning-noise

Recent Developments in Supervised Learning With Noise. The Simons Institute for the Theory of Computing, Dec. 18, 2020, https://simons.berkeley.edu/talks/recent-developments-supervised-learning-noise 

          @misc{ scivideos_16891,
            doi = {},
            url = {https://simons.berkeley.edu/talks/recent-developments-supervised-learning-noise},
            author = {},
            keywords = {},
            language = {en},
            title = {Recent Developments in Supervised Learning With Noise},
            publisher = {The Simons Institute for the Theory of Computing},
            year = {2020},
            month = {dec},
            note = {16891 see, \url{https://scivideos.org/Simons-Institute/16891}}
          }
Ilias Diakonikolas (UW Madison)
Talk number16891
Source RepositorySimons Institute
Subject

Abstract

The talk will survey recent progress on the computational complexity of binary classification in the presence of benign label noise. In particular, we will will give an overview of the key ideas behind known algorithms and computational hardness results for learning halfspaces (and other concept classes) in both the distribution free and the distribution specific PAC model.