Shallow2Deep: Restraining Neural Networks Opacity through Neural Architecture Search


Andrea Agiollo, Giovanni Ciatto, Andrea Omicini  

Recently, the Deep Learning (DL) research community has focused on developing efficient and highly performing Neural Networks (NN). Meanwhile, the eXplainable AI (XAI) research community has focused on making Machine Learning (ML) and Deep Learning methods interpretable and transparent, seeking explainability. This work is a preliminary study on the applicability of Neural Architecture Search (NAS) (a sub-field of DL looking for automatic design of NN structures) in XAI. We propose Shallow2Deep, an evolutionary NAS algorithm that exploits local variability to restrain opacity of DL-systems through NN architectures simplification. Shallow2Deep effectively reduces NN complexity – therefore their opacity – while reaching state-of-the-art performances. Unlike its competitors, Shallow2Deep promotes variability of localised structures in NN, helping to reduce NN opacity. The proposed work analyses the role of local variability in NN architectures design, presenting experimental results that show how this feature is actually desirable.

(keywords) Neural Architecture Search; Evolutionary Algorithm; Opacity; Interpretability

Explainable and Transparent AI and Multi-Agent Systems. Third International Workshop, EXTRAAMAS 2021, Virtual Event, May 3–7, 2021, Revised Selected Papers, Lecture Notes in Computer Science (Lecture Notes in Artificial Intelligence) 12688, pages 63-82,  2021, Springer Nature, Basel, Switzerland.

@incollection{shallow2deep-extraamas2021,
address = {Basel, Switzerland},
author = {Agiollo, Andrea and Ciatto, Giovanni and Omicini, Andrea},
booktitle = {Explainable and Transparent AI and Multi-Agent Systems. Third International Workshop, EXTRAAMAS 2021, Virtual Event, May 3--7, 2021, Revised Selected Papers},
doi = {10.1007/978-3-030-82017-6_5},
editor = {Calvaresi, Davide and Najjar, Amro and Winikoff, Michael and Fr{\"a}mling, Kary},
isbn = {978-3-030-82016-9},
isbn-online = {978-3-030-82017-6},
issn = {0302-9743},
keywords = {Neural Architecture Search; Evolutionary Algorithm; Opacity; Interpretability},
pages = {63--82},
publisher = {Springer Nature},
series = {Lecture Notes in Computer Science},
subseries = {Lecture Notes in Artificial Intelligence},
title = {{\it Shallow2Deep}: Restraining Neural Networks Opacity through Neural Architecture Search},
url = {http://link.springer.com/10.1007/978-3-030-82017-6_5},
volume = 12688,
year = 2021}

    

Publication

Shallow2Deep: Restraining Neural Networks Opacity through Neural Architecture Search

— status

published  

— authors

Andrea Agiollo, Giovanni Ciatto, Andrea Omicini

URLs & IDs

original page

— DOI

10.1007/978-3-030-82017-6_5

— DBLP

conf/atal/AgiolloCO21

— print ISSN

0302-9743

— online ISSN

1611-3349

— print ISBN

978-3-030-82016-9

— online ISBN

978-3-030-82017-6

BibTeX

— BibTeX ID
shallow2deep-extraamas2021
— BibTeX category
incollection

APICe

— event
EXTRAAMAS 2021
— journal/series
LNCS
— talk
Shallow2Deep: Restraining Neural Networks Opacity through Neural Architecture Search

Partita IVA: 01131710376 - Copyright © 2008-2021 APICe@DISI Research Group - PRIVACY