BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:US/Eastern
BEGIN:DAYLIGHT
DTSTART:20170312T030000
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
RRULE:FREQ=YEARLY;BYDAY=2SU;BYMONTH=3
TZNAME:EDT
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20171105T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
RRULE:FREQ=YEARLY;BYDAY=1SU;BYMONTH=11
TZNAME:EST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20171004T191406Z
UID:D7B0AD0D-A2EB-11E7-A02F-0050568D7F66
DTSTART;TZID=US/Eastern:20171010T183000
DTEND;TZID=US/Eastern:20171010T203000
DESCRIPTION:Deep neural networks have led to dramatic improvements in perfo
 rmance for many machine learning tasks\, yet the mathematical reasons for 
 this success remain largely unclear. In this talk we present recent develo
 pments in the mathematical framework of convolutive neural networks (CNN).
  In particular we discuss the scattering network of Mallat and how it rela
 tes to another problem in harmonic analysis\, namely the phase retrieval p
 roblem. Then we discuss the general convolutive neural network from a theo
 retician point of view. We present Lipschitz analysis results using two an
 alytical methods: the chain rule (or backpropagation) and the storage func
 tion method inspired by Mallat's scattering network analysis. Towards the 
 end of the talk we discuss how these theoretical results can be applied in
  practice\, and in particular we mention various design methods that incor
 porate Lipschitz bounds as penalty terms into optimization problems.\n\nCo
 -sponsored by: WASH Signal Processing Society\n\nSpeaker(s): Prof. Radu Ba
 lan\, \, Prof. Radu Balan\, \, Prof. Radu Balan\, \n\nAgenda: \nLight dinn
 er and refreshments will be served at 6:30pm\; lecture will start at 7pm.\
 n\nIf arriving by car\, feel free to use parking lot 11b\, which is open f
 or general access after 4pm. If arriving by Metro\, free shuttle service i
 s available from the College Park station to campus.\n\nRoom: 2460 (ECE De
 partment's Colloquium Room)\, Bldg: AV Williams - Parking Lot 11B\, Univer
 sity of Maryland\, College Park\, Maryland\, United States
LOCATION:Room: 2460 (ECE Department's Colloquium Room)\, Bldg: AV Williams 
 - Parking Lot 11B\, University of Maryland\, College Park\, Maryland\, Uni
 ted States
ORGANIZER:joel.i.goodman@ieee.org
SEQUENCE:13
SUMMARY:When Harmonic Analysis meets Machine Learning: Lipschitz Analysis o
 f Deep Convolution Networks 
URL;VALUE=URI:https://events.vtools.ieee.org/m/47297
X-ALT-DESC:Description: <br /><p><span style="color: #000000\;">Deep neural
  networks have led to dramatic improvements in performance for many machin
 e learning tasks\, yet the mathematical reasons for this success remain la
 rgely unclear. In this talk we present recent developments in the mathemat
 ical framework of convolutive neural networks (CNN). In particular we disc
 uss the scattering network of Mallat and how it relates to another problem
  in harmonic analysis\, namely the phase retrieval problem. Then we discus
 s the general convolutive neural network from a theoretician point of view
 . We present Lipschitz analysis results using two analytical methods: the 
 chain rule (or backpropagation) and the storage function method inspired b
 y Mallat's scattering network analysis. Towards the end of the talk we dis
 cuss how these theoretical results can be applied in practice\, and in par
 ticular we mention various design methods that incorporate Lipschitz bound
 s as penalty terms into optimization problems.</span></p><br /><br />Agend
 a: <br /><p>Light dinner and refreshments will be served at 6:30pm\; lectu
 re will start at 7pm.&nbsp\;</p>\n<p>If arriving by car\, feel free to use
  parking lot 11b\, which is open for general access after 4pm. If arriving
  by Metro\, free shuttle service is available from the College Park statio
 n to campus.&nbsp\;</p>
END:VEVENT
END:VCALENDAR