Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Evolving programs to build artificial neural networks

Evolving programs to build artificial neural networks Evolving Programs to Build Artificial Neural Networks Dennis G. Wilson, University of Toulouse; Toulouse Mind & Brain Institute, Toulouse, France Introduction The biological brain has inspired many artificial intelligence algorithms, especially artificial neural networks (ANNs). Originally designed to replicate neural functionality [10], these models are now used in a wide variety of applications and have shown impressive ability in computer vision, natural language processing, and control. Despite these significant advances, contemporary ANNs are far removed from their biological sources and do not demonstrate the same ability to generalize to new tasks and in new environments as biological organisms [4, 7]. We suggest that the focus on synaptic weight change as the learning mechanisms in ANNs is a contributing factor to their limitations in generalization. As neural development, which changes network structure, is an integral part of biological learning, we propose that structural development should be included in learning in ANNs. Developmental neural networks have not widely been explored in the literature and there remains a need for a concerted effort to explore a greater variety of effective models. In this article, we summarize a conceptually simple neural model named IMPROBED. This model represents development with two programs that control http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ACM SIGEVOlution Association for Computing Machinery

Evolving programs to build artificial neural networks

ACM SIGEVOlution , Volume 15 (1): 6 – Apr 20, 2022

Loading next page...
 
/lp/association-for-computing-machinery/evolving-programs-to-build-artificial-neural-networks-yQ0z9l8koh

References (93)

Publisher
Association for Computing Machinery
Copyright
Copyright © 2022 Copyright is held by the owner/author(s)
ISSN
1931-8499
eISSN
1931-8499
DOI
10.1145/3532942.3532944
Publisher site
See Article on Publisher Site

Abstract

Evolving Programs to Build Artificial Neural Networks Dennis G. Wilson, University of Toulouse; Toulouse Mind & Brain Institute, Toulouse, France Introduction The biological brain has inspired many artificial intelligence algorithms, especially artificial neural networks (ANNs). Originally designed to replicate neural functionality [10], these models are now used in a wide variety of applications and have shown impressive ability in computer vision, natural language processing, and control. Despite these significant advances, contemporary ANNs are far removed from their biological sources and do not demonstrate the same ability to generalize to new tasks and in new environments as biological organisms [4, 7]. We suggest that the focus on synaptic weight change as the learning mechanisms in ANNs is a contributing factor to their limitations in generalization. As neural development, which changes network structure, is an integral part of biological learning, we propose that structural development should be included in learning in ANNs. Developmental neural networks have not widely been explored in the literature and there remains a need for a concerted effort to explore a greater variety of effective models. In this article, we summarize a conceptually simple neural model named IMPROBED. This model represents development with two programs that control

Journal

ACM SIGEVOlutionAssociation for Computing Machinery

Published: Apr 20, 2022

There are no references for this article.