This thesis aims to analyze and optimize convolutional neural networks using social network analysis techniques. After an overview of Neural Networks and Social Network Analysis, a general method is proposed for the definition of a Social Network starting from a Convolutional Neural Network. Subsequently an example of application of the introduced method is reported, using a known Convolutional Neural Network, that is the VGG16, pre-trained on the CIFAR10 dataset. Eventually, the networks obtained are analyzed through the use of Social Network Analysis techniques, with the aim of deducing information and making any optimizations of the starting network. The thesis ends with an analysis of the results obtained, aimed at identifying the limits of the approach used, and with an overview of some possible future developments aimed at overcoming these limits.
La presente tesi ha come scopo l'analisi e l’ottimizzazione delle Reti Neurali Convoluzionali mediante le tecniche di Social Network Analysis. Dopo una panoramica sulle Reti Neurali e sulla Social Network Analysis, viene proposto un metodo generale per la definizione di una Rete Sociale a partire da una Rete Neurale Convoluzionale. Successivamente viene riportato un esempio di applicazione del metodo introdotto, utilizzando una Rete Neurale Convoluzionale nota, ovvero la VGG16, pre-allenata sul dataset CIFAR10. Le reti così ottenute vengono, infine, analizzate mediante l'utilizzo di tecniche di Social Network Analysis, con lo scopo di dedurre informazioni ed effettuare eventuali ottimizzazioni della rete di partenza. La tesi si conclude con un'analisi dei risultati ottenuti, volta ad individuare i limiti dell'approccio utilizzato, e con una panoramica su alcuni possibili sviluppi futuri volti a superare tali limiti.
Rappresentazione ed analisi di Reti Neurali Convoluzionali attraverso la Social Network Analysis
DI MASSIMO, PAOLO
2019/2020
Abstract
This thesis aims to analyze and optimize convolutional neural networks using social network analysis techniques. After an overview of Neural Networks and Social Network Analysis, a general method is proposed for the definition of a Social Network starting from a Convolutional Neural Network. Subsequently an example of application of the introduced method is reported, using a known Convolutional Neural Network, that is the VGG16, pre-trained on the CIFAR10 dataset. Eventually, the networks obtained are analyzed through the use of Social Network Analysis techniques, with the aim of deducing information and making any optimizations of the starting network. The thesis ends with an analysis of the results obtained, aimed at identifying the limits of the approach used, and with an overview of some possible future developments aimed at overcoming these limits.File | Dimensione | Formato | |
---|---|---|---|
Tesi definitiva pdfa.pdf
Open Access dal 20/02/2023
Dimensione
3.12 MB
Formato
Adobe PDF
|
3.12 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/4642