2026-09-18T20:17:37Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/113893 2023-07-11 col_2142_5131 col_2142_8888 com_2142_5130 com_2142_8887 com_2142_234

Srikant, Rayadurgam Srikant, Rayadurgam Viswanath, Pramod Raginsky, Maxim Sun, Ruoyu Lee, Jason D. 2022-04-29T21:34:46Z 2021-12 Submission original under an indefinite embargo labeled 'Open Access'. The submission was exported from vireo on 2022-04-06 without embargo terms The student, Shiyu Liang, accepted the attached license on 2021-12-01 at 11:06. The student, Shiyu Liang, submitted this Dissertation for approval on 2021-12-01 at 11:06. This Dissertation was approved for publication on 2021-12-01 at 16:30. DSpace SAF Submission Ingestion Package generated from Vireo submission #17333 on 2022-04-06 at 17:10:34 Made available in DSpace on 2022-04-29T21:34:46Z (GMT). No. of bitstreams: 3 LIANG-DISSERTATION-2021.pdf: 658537 bytes, checksum: e0c47e791cae3180b0683ed2aee50f29 (MD5) LICENSE.txt: 4208 bytes, checksum: 7a8a3f092c6bdf10766efc8ce2f5049b (MD5) PROQUEST_LICENSE.txt: 4554 bytes, checksum: 020958fc887f59bb41f74a25430b239e (MD5) Previous issue date: 2021-12-01 application/pdf en text The role of explicit regularization in overparameterized neural networks Liang, Shiyu 2021-12-01 Engineering 2022-04-29T21:34:46Z Recent theoretical works on over-parameterized neural nets have focused on two aspects: optimization and generalization. Many existing works that study optimization and generalization together are based on the neural tangent kernel and require a very large width. In this dissertation, we are interested in the following two questions: for a binary classification problem with two-layer mildly over-parameterized ReLU network, (1) does every local minimum memorize and generalize well? and (2) can we find a set of parameters that result in small test error in polynomial time? We first show that the landscape of loss functions with explicit regularization has the following property: all local minima, and certain other points which are only stationary in certain directions, achieve small test error. We then prove that, for convolutional neural nets, there is an algorithm which finds one of these points in polynomial time (in the input dimension and the number of data points). In addition, we prove that for a fully connected neural net, with an additional assumption on the data distribution, there is a polynomial-time algorithm to find one of these points. Thesis eng http://hdl.handle.net/2142/113893 Copyright 2021 Shiyu Liang Electrical & Computer Eng Electrical & Computer Engr University of Illinois at Urbana-Champaign Dissertation Ph.D.