2026-09-18T19:36:31Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/92687 2023-07-11 col_2142_5131 com_2142_5130

DSpace SAF Submission Ingestion Package generated from Vireo submission #9397 on 2016-11-09 at 10:19:06 Made available in DSpace on 2016-11-10T17:49:18Z (GMT). No. of bitstreams: 2 YEH-THESIS-2016.pdf: 623799 bytes, checksum: 4ba29b019a9f5fb40582880b0e22c9ff (MD5) LICENSE.txt: 4208 bytes, checksum: 3e6ff98c0e570cf11f447ad1722d50ca (MD5) Previous issue date: 2016-04-26 application/pdf http://hdl.handle.net/2142/92687 en Copyright 2016 Raymond Yeh convolutional neural network deep learning Stable and symmetric convolutional neural network text text Do, Minh N. Yeh, Raymond Alexander 2016-11-10T17:49:18Z 2016-11-10T17:49:18Z 2016-04-26 2016-08 First we present a proof that convolutional neural networks (CNNs) with max-norm regularization, max-pooling, and Relu non-linearity are stable to additive noise. Second, we explore the use of symmetric and antisymmetric filters in a baseline CNN model on digit classification, which enjoys the stability to additive noise. Experimental results indicate that the symmetric CNN outperforms the baseline model for nearly all training sizes and matches the state-of-the-art deep-net in the cases of limited training examples. Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2016-11-09 without embargo terms The student, Raymond Yeh, accepted the attached license on 2016-04-24 at 10:31. The student, Raymond Yeh, submitted this Thesis for approval on 2016-04-24 at 10:36. This Thesis was approved for publication on 2016-04-26 at 11:09. Electrical & Computer Eng Electrical & Computer Engr University of Illinois at Urbana-Champaign Thesis M.S.