Adapting the Size of Artificial Neural Networks Using Dynamic Auto-Sizing
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21240%2F22%3A00363052" target="_blank" >RIV/68407700:21240/22:00363052 - isvavai.cz</a>
Result on the web
<a href="https://ieeexplore.ieee.org/abstract/document/10000471" target="_blank" >https://ieeexplore.ieee.org/abstract/document/10000471</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1109/CSIT56902.2022.10000471" target="_blank" >10.1109/CSIT56902.2022.10000471</a>
Alternative languages
Result language
angličtina
Original language name
Adapting the Size of Artificial Neural Networks Using Dynamic Auto-Sizing
Original language description
We introduce dynamic auto-sizing, a novel approach to training artificial neural networks which allows the models to automatically adapt their size to the problem domain. The size of the models can be further controlled during the learning process by modifying the applied strength of regularization. The ability of dynamic auto-sizing models to expand or shrink their hidden layers is achieved by periodically growing and pruning entire units such as neurons or filters. For this purpose, we introduce weighted L1 regularization, a novel regularization method for inducing structured sparsity. Besides analyzing the behavior of dynamic auto-sizing, we evaluate predictive performance of models trained using the method and show that such models can provide a predictive advantage over traditional approaches.
Czech name
—
Czech description
—
Classification
Type
D - Article in proceedings
CEP classification
—
OECD FORD branch
10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Result continuities
Project
—
Continuities
S - Specificky vyzkum na vysokych skolach
Others
Publication year
2022
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Article name in the collection
IEEE 17th International Conference on Computer Science and Information Technologies
ISBN
979-8-3503-3431-9
ISSN
—
e-ISSN
—
Number of pages
5
Pages from-to
592-596
Publisher name
IEEE
Place of publication
Dortmund
Event location
Lvov
Event date
Nov 10, 2022
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
000927642900139