BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Mad Max: Affine Spline Insights into Deep Learning
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260912T230944Z
UID:tag:localist.com\,2008:EventInstance_34536580078534
DTSTART:20201023T190000Z
DTEND:20201023T201500Z
DESCRIPTION:Please register here to attend the EE Seminars via Zoom this fa
 ll. Once you register\, you will receive the Zoom link. You only need to r
 egister once to be able to attend any of the seminars. \n\n---\n\nWe build
  a rigorous bridge between deep networks (DNs) and approximation theory vi
 a spline functions and operators. Our key result is that a large class of 
 DNs can be written as a composition of max-affine spline operators (MASOs)
 \, which provide a powerful portal through which to view and analyze their
  inner workings. For instance\, conditioned on the input signal\, the outp
 ut of a MASO DN can be written as a simple affine transformation of the in
 put. This implies that a DN constructs a set of signal-dependent\, class-s
 pecific templates against which the signal is compared via a simple inner 
 product\; we explore the links to the classical theory of optimal classifi
 cation via matched filters and the effects of data memorization. The splin
 e partition of the input signal space that is implicitly induced by a MASO
  directly links DNs to the theory of vector quantization (VQ) and K-means 
 clustering\, which opens up new geometric avenue to study how DNs organize
  signals in a hierarchical and multiscale fashion.
LOCATION:Zoom
SUMMARY:Mad Max: Affine Spline Insights into Deep Learning
URL;VALUE=URI:https://events.seas.harvard.edu/event/guest_speaker_richard_b
 araniuk
CATEGORIES:Colloquia / Seminar / Lecture
END:VEVENT
END:VCALENDAR