BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Romain Couillet (CentraleSupélec\, Université Paris-Saclay) DTSTART:20210507T083000Z DTEND:20210507T100000Z DTSTAMP:20260423T022715Z UID:MEGA/25 DESCRIPTION:Title: Wh y Random Matrices can Change the Future of Research in AI?\nby Romain Couillet (CentraleSupélec\, Université Paris-Saclay) as part of Séminai re MEGA\n\n\nAbstract\nMachine learning and AI algorithms are becoming inc reasingly more powerful but also increasingly more complex\, mathematicall y less tractable\, and energetically less environmental friendly. In this talk\, we will demonstrate that large dimensional statistics\, and particu larly random matrix theory\, simultaneously (i) explains why ML algorithms are so stable when dealing with large dimensional data\, (ii) manages to break the difficulties that make these algorithms mathematically intractab le (non-linearities and data modelling)\, thereby (iii) allowing for the f irst time to get (iii-a) an inside understanding of the algorithms\, of th eir multiple biases and\, most crucially\, of their quite counter-intuitiv e behavior as well as (iii-b) a toolbox to easily improve the algorithms p erformance and cost efficiency. Possibly even more surprisingly\, the univ ersality notion in random matrix theory shows (iv) why ML algorithms appli ed to intricate real data (in general impossible to model) behave the same as when applied to elementary Gaussian random vector models.\n\nThe cours e will introduce basic notions of random matrix theory by emphasizing on t he counter-intuitive behavior of large dimensional data (so to raise aware ness in the audience). These notions will be applied to a range of telling applications in machine learning (spectral clustering\, semi-supervised l earning\, transfer learning\, low-cost processing\, etc.).\n\nThe audience can dynamically decide on which topic they'd like me to cover preferably. A time for debate will also be given for the audience to react on the pre sentation. An extensive coverage of the class material is available online in the upcoming book “Romain COUILLET\, Zhenyu LIAO\, “Random Matrix Theory for Machine Learning” https://romaincouillet.hebfree.org/book.htm l\n LOCATION:https://researchseminars.org/talk/MEGA/25/ END:VEVENT END:VCALENDAR