BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Mathematics Seminar: Functional Dimension in ReLU Neural Netwo
 rks
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260812T142457Z
UID:tag:localist.com\,2008:EventInstance_51887873581863
DTSTART:20260128T210000Z
DTEND:20260128T220000Z
DESCRIPTION:Presenter:     Dr. Christopher Cornwell\, Department of Mathema
 tics\, Towson University\n\nAbstract:      Associated to the parameters of
  a neural network with a ReLU activation function is a number called the f
 unctional dimension.  Roughly speaking\, the functional dimension measures
  the number of degrees of freedom for determining a new network function b
 y perturbing the parameters\, as is done during network training.  For a c
 hoice of network and parameters\, a tight upper bound on the functional di
 mension is known – one that is strictly less than the number of paramete
 rs\; conditions under which the functional dimension is strictly less than
  that upper bound have been explored\, and it is an active area of researc
 h.  In this talk\, after an introduction to the ideas above\, I will discu
 ss contributions of me and collaborators to this area.  Several recent wor
 ks by others have focused on the existence of a positive measure subset of
  parameters that achieve the upper bound.  Our work explores the probabili
 ty of getting initial parameters that fail to achieve the upper bound\, us
 ing standard assumptions on initial parameters as random variables.
GEO:39.390591;-76.605888
LOCATION:7800 York Road\, Room 321
SUMMARY:Mathematics Seminar: Functional Dimension in ReLU Neural Networks
URL;VALUE=URI:https://events.towson.edu/event/mathematics-seminar-functiona
 l-dimension-in-relu-neural-networks
CATEGORIES:Academics
CATEGORIES:Academic Seminar
END:VEVENT
END:VCALENDAR