ICTS:32491

Turing Lecture: Overparametrized models: linear theory and its limits

(2025). Turing Lecture: Overparametrized models: linear theory and its limits. SciVideos. https://scivideos.org/icts-tifr/32491

Turing Lecture: Overparametrized models: linear theory and its limits. SciVideos, Aug. 12, 2025, https://scivideos.org/icts-tifr/32491 

          @misc{ scivideos_ICTS:32491,
            doi = {},
            url = {https://scivideos.org/icts-tifr/32491},
            author = {},
            keywords = {},
            language = {en},
            title = {Turing Lecture: Overparametrized models: linear theory and its limits},
            publisher = {},
            year = {2025},
            month = {aug},
            note = {ICTS:32491 see, \url{https://scivideos.org/icts-tifr/32491}}
          }
Andrea Montanari
Talk numberICTS:32491
Source RepositoryICTS-TIFR

Abstract

The success of modern AI models defies classical theoretical wisdom. Classical theory recommended the use of convex optimization, and yet AI models learn by optimizing highly non-convex function. Classical theory prescribed to control model complexity and yet AI models are very complex, so complex that they often memorize the training data. Classical wisdom recommends a careful and interpretable choice of model architecture, and yet modern architectures rarely offer a parsimonious representation of a target distribution class.

The discovery that learning can take place in completely unexpected scenario poses beautiful conceptual challenges. I will try to survey recent work towards addressing them.