[Colloquium] Fraser/MS Presentation/Nov 8, 2011

[Margaret Jaffey]

This is an announcement of Maia Fraser's MS Presentation.

------------------------------------------------------------------------------
Date:  Tuesday, November 8, 2011

Time:  9:00 AM

Place:  Ryerson 251

M.S. Candidate:  Maia Fraser

M.S. Paper Title: Structural Observations on Neural Networks for
Hierarchically Derived Reproducing Kernels

Abstract:
This thesis addresses structural aspects of neural networks (NN's)
used for producing hierarchically derived reproducing kernels in the
CBCL model. It also shows that convolution neural networks (CNN's) may
be viewed as a special case of a generalized model we call the
extended CBCL model. The main results of the thesis use Hilbert space
geometry to show that for NN's in the extended CBCL model, the use of
multiple layers and/or non-linear neural responses does not broaden
the class of kernels computed, rather at issue seems to be the amount
of effort involved in specializing the NN to its intended function.
More precisely, the thesis shows how various design parameters of the
NN (transformation sets, neural responses, templates, depth) may be
traded off against each other without changing the map computed by the
NN.

Maia's advisor is Prof. Yali Amit

Login to the Computer Science Department website for details:
 https://www.cs.uchicago.edu/phd/ms_announcements#maia

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Margaret P. Jaffey            margaret at cs.uchicago.edu
Department of Computer Science
Student Support Rep (Ry 156)               (773) 702-6011
The University of Chicago      http://www.cs.uchicago.edu
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=