On the representativeness of convolutional neural networks layers
Document typePart of book or chapter of book
PublisherIOS PRESS EBOOKS
Rights accessOpen Access
Convolutional Neural Networks (CNN) are the most popular of deep network models due to their applicability and success in image processing. Although plenty of effort has been made in designing and training better discriminative CNNs, little is yet known about the internal features these models learn. Questions like, what specific knowledge is coded within CNN layers, and how can it be used for other purposes besides discrimination, remain to be answered. To advance in the resolution of these questions, in this work we extract features from CNN layers, building vector representations from CNN activations. The resultant vector embedding is used to represent first images and then known image classes. On those representations we perform an unsupervised clustering process, with the goal of studying the hidden semantics captured in the embedding space. Several abstract entities untaught to the network emerge in this process, effectively defining a taxonomy of knowledge as perceived by the CNN. We evaluate and interpret these sets using WordNet, while studying the different behaviours exhibited by the layers of a CNN model according to their depth. Our results indicate that, while top (i.e., deeper) layers provide the most representative space, low layers also define descriptive dimensions.
CitationGarcía, D., Moreno, J., Ramos-Pollan, R., Barrios, R., Béjar, J., Cortés, C., Ayguadé, E., Labarta, J., Suzumura, T. On the representativeness of convolutional neural networks layers. A: "Artificial Intelligence Research and Development: proceedings of the 19th International Conference of the Catalan Association for Artificial Intelligence: Barcelona, Catalonia, Spain, October 19–21, 2016". IOS PRESS EBOOKS, 2016, p. 29-38.
- KEMLG - Grup d'Enginyeria del Coneixement i Aprenentatge Automàtic - Capítols de llibre 
- CAP - Grup de Computació d'Altes Prestacions - Capítols de llibre 
- Computer Sciences - Capítols de llibre 
- Departament de Ciències de la Computació - Capítols de llibre 
- Departament d'Arquitectura de Computadors - Capítols de llibre 
All rights reserved. This work is protected by the corresponding intellectual and industrial property rights. Without prejudice to any existing legal exemptions, reproduction, distribution, public communication or transformation of this work are prohibited without permission of the copyright holder