Phylogenetic replay learning in deep neural networks
Joint Authors
Akdag, Herman
Glafkides, Jean Patrice
Sher, Gene I.
Source
Jordanian Journal of Computetrs and Information Technology
Issue
Vol. 8, Issue 3 (30 Sep. 2022), pp.218-231, 14 p.
Publisher
Princess Sumaya University for Technology
Publication Date
2022-09-30
Country of Publication
Jordan
No. of Pages
14
Main Subjects
Information Technology and Computer Science
Abstract EN
Though substantial advancements have been made in training deep neural networks, one problem remains, the vanishing gradient.
The very strength of deep neural networks, their depth, is also unfortunately their problem, due to the difficulty of thoroughly training the deeper layers due to the vanishing gradient.
This paper proposes "Phylogenetic Replay Learning", a learning methodology that substantially alleviates the vanishing-gradient problem.
Unlike the residual learning methods, it does not restrict the structure of the model.
Instead, it leverages elements from neuroevolution, transfer learning and layer-by-layer training.
We demonstrate that this new approach is able to produce a better performing model and by calculating Shannon entropy of weights, we show that the deeper layers are trained much more thoroughly and contain statistically significantly more information than when a model is trained in a traditional brute force manner.
American Psychological Association (APA)
Glafkides, Jean Patrice& Sher, Gene I.& Akdag, Herman. 2022. Phylogenetic replay learning in deep neural networks. Jordanian Journal of Computetrs and Information Technology،Vol. 8, no. 3, pp.218-231.
https://search.emarefa.net/detail/BIM-1435894
Modern Language Association (MLA)
Glafkides, Jean Patrice…[et al.]. Phylogenetic replay learning in deep neural networks. Jordanian Journal of Computetrs and Information Technology Vol. 8, no. 3 (Sep. 2022), pp.218-231.
https://search.emarefa.net/detail/BIM-1435894
American Medical Association (AMA)
Glafkides, Jean Patrice& Sher, Gene I.& Akdag, Herman. Phylogenetic replay learning in deep neural networks. Jordanian Journal of Computetrs and Information Technology. 2022. Vol. 8, no. 3, pp.218-231.
https://search.emarefa.net/detail/BIM-1435894
Data Type
Journal Articles
Language
English
Notes
Includes bibliographical references : p. 230-231
Record ID
BIM-1435894