<?xml version="1.0" encoding="UTF-8"?><?xml-stylesheet type="text/xsl" href="static/style.xsl"?><OAI-PMH xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/ http://www.openarchives.org/OAI/2.0/OAI-PMH.xsd"><responseDate>2026-11-22T19:03:12Z</responseDate><request verb="GetRecord" identifier="oai:uwspace.uwaterloo.ca:10012/22649" metadataPrefix="dim">https://dspacemainprd01.lib.uwaterloo.ca/server/oai/request</request><GetRecord><record><header><identifier>oai:uwspace.uwaterloo.ca:10012/22649</identifier><datestamp>2026-01-20T08:03:18Z</datestamp><setSpec>com_10012_9900</setSpec><setSpec>com_10012_1</setSpec><setSpec>col_10012_9914</setSpec><setSpec>col_10012_6</setSpec></header><metadata><dim:dim xmlns:dim="http://www.dspace.org/xmlns/dspace/dim" xmlns:doc="http://www.lyncode.com/xoai" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.dspace.org/xmlns/dspace/dim http://www.dspace.org/schema/dim.xsd">
   <dim:field mdschema="dc" element="contributor" qualifier="author">Mokadem, Frank</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="accessioned">2025-11-26T16:42:28Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="available">2025-11-26T16:42:28Z</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="issued">2025-11-26</dim:field>
   <dim:field mdschema="dc" element="date" qualifier="submitted">2025-11-19</dim:field>
   <dim:field mdschema="dc" element="identifier" qualifier="uri">https://hdl.handle.net/10012/22649</dim:field>
   <dim:field mdschema="dc" element="description" qualifier="abstract">Convolutional Neural Network (CNN) remain the architecture of choice for computer vision tasks on compute-constrained platforms such as edge and personal devices, delivering both close to state-of-the-art performance metrics and linear inference complexity with respect to input resolution and number of channels. However, the deployment of larger and more complex CNN architectures is limited by the restrained memory offered by such platforms. This brings about a need to compress pretrained CNN into smaller models in number of parameters while controlling for degradation in performance. 
This thesis tackles CNN compression using low rank approximation of convolution layers using Tucker Decomposition (TD). We introduce a new heuristics-based Neural Architectural Search procedure to select low rank configurations for the convolution tensors, which we call Heterogeneous Tucker Decomposition (HTD). 

Standard low rank approximation using TD factorizes and approximates convolution layers using uniform ranks for all convolution tensors, then applies a few fine–tuning epochs to recover degradation in performance. An approach we show to be suboptimal against a heterogeneous selection of ranks for each convolution layer, followed by same number of fine-tuning epochs.

Our primary contribution is the development and evaluation of TD, which applies layers-pecific compression rate (low rank divided by full rank) inferred from a Neural Architectural Search (NAS) process. Furthermore, we introduce a sampling heuristic to efficiently
explore the search space of layer-specific compression rates, thus preserving performance
while significantly reducing search time.

We present a mathematical formulation for the HTD optimization problem and an NAS algorithm to find admissible solutions. We test our approach on multiple varieties of CNN architectures: AlexNet, VGG16, and ResNet18, adapted for the MNIST classification task. Our findings confirm that HTD performs better than TD on all models tested. For the same compression rate, HTD enables to recover a higher precision after fine-tuning, with gains ranging from 1.2% to 5.8%. For equivalent accuracy targets, HTD delivers 15-30% higher compression rates than TD.

This thesis advances Neural Architectural Search by highlighting the efficacy of heterogeneous tensor decomposition approaches. It provides a robust framework for their implementation and evaluation, with significant implications for deploying convolutional deep learning models in resource-limited settings. Future work will explore incorporating low-rank constraints as a regularization objective during training, potentially enabling end-to-end compression-aware optimization.</dim:field>
   <dim:field mdschema="dc" element="language" qualifier="iso">en</dim:field>
   <dim:field mdschema="dc" element="publisher" lang="en">University of Waterloo</dim:field>
   <dim:field mdschema="dc" element="subject">CNN</dim:field>
   <dim:field mdschema="dc" element="subject">Tucker decomposition</dim:field>
   <dim:field mdschema="dc" element="subject">neural architectural search</dim:field>
   <dim:field mdschema="dc" element="title">Heterogeneous Decomposition of Convolutional Neural Networks Using Tucker Decomposition</dim:field>
   <dim:field mdschema="dc" element="type">Master Thesis</dim:field>
   <dim:field mdschema="dc" element="pending">false</dim:field>
   <dim:field mdschema="uws-etd" element="degree" qualifier="department">Systems Design Engineering</dim:field>
   <dim:field mdschema="uws-etd" element="degree" qualifier="discipline">System Design Engineering</dim:field>
   <dim:field mdschema="uws-etd" element="degree" qualifier="grantor" lang="en">University of Waterloo</dim:field>
   <dim:field mdschema="uws-etd" element="degree">Master of Applied Science</dim:field>
   <dim:field mdschema="uws-etd" element="embargo" qualifier="terms">0</dim:field>
   <dim:field mdschema="uws" element="contributor" qualifier="advisor">Alexander, Wong</dim:field>
   <dim:field mdschema="uws" element="contributor" qualifier="affiliation1">Faculty of Engineering</dim:field>
   <dim:field mdschema="uws" element="published" qualifier="city" lang="en">Waterloo</dim:field>
   <dim:field mdschema="uws" element="published" qualifier="country" lang="en">Canada</dim:field>
   <dim:field mdschema="uws" element="published" qualifier="province" lang="en">Ontario</dim:field>
   <dim:field mdschema="uws" element="typeOfResource" lang="en">Text</dim:field>
   <dim:field mdschema="uws" element="peerReviewStatus" lang="en">Unreviewed</dim:field>
   <dim:field mdschema="uws" element="scholarLevel" lang="en">Graduate</dim:field>
   <dim:field mdschema="others" element="access-status">open.access</dim:field>
</dim:dim>
</metadata></record></GetRecord></OAI-PMH>