Deriving composite functions as activation function for neural networks or for wavenets
Other Title(s)
اشتقاق دالة مركبة كدالة تحفيز للشبكات العصبية أو في شبكات المويجات
Author
Source
al-Qadisiyah Journal for Computer Science and Mathematics
Issue
Vol. 6, Issue 2 (31 Dec. 2014), pp.19-29, 11 p.
Publisher
University of al-Qadisiyah College of computer Science and Information Technology
Publication Date
2014-12-31
Country of Publication
Iraq
No. of Pages
11
Main Subjects
Abstract EN
This study present new idea in way to more future wonderful ideas, exploiting mathematical side of specific, through several ways employing very wide resources of mathematical functions and their properties that with natural in improve features of activation function, from these studies this study which drive an activation function through compose two (and more) functions with properties to be continuous and drivable.
We can drive membership function under fuzzy logic by this as composite function on wavelet function or vice versa, and for many different functions (as future work).
Also there idea to compose wavelets from families such SLOG mother wavelets with RASP or POLYWOG wavelets, that specifically arises from NNs representation and classification problems ,which also drivable .Each one from these functions have derivative which also can be composed to get an other activation function .
American Psychological Association (APA)
Ulaywi, Bushra Husayn. 2014. Deriving composite functions as activation function for neural networks or for wavenets. al-Qadisiyah Journal for Computer Science and Mathematics،Vol. 6, no. 2, pp.19-29.
https://search.emarefa.net/detail/BIM-795912
Modern Language Association (MLA)
Ulaywi, Bushra Husayn. Deriving composite functions as activation function for neural networks or for wavenets. al-Qadisiyah Journal for Computer Science and Mathematics Vol. 6, no. 2 (2014), pp.19-29.
https://search.emarefa.net/detail/BIM-795912
American Medical Association (AMA)
Ulaywi, Bushra Husayn. Deriving composite functions as activation function for neural networks or for wavenets. al-Qadisiyah Journal for Computer Science and Mathematics. 2014. Vol. 6, no. 2, pp.19-29.
https://search.emarefa.net/detail/BIM-795912
Data Type
Journal Articles
Language
English
Notes
Includes bibliographical references : p. 29
Record ID
BIM-795912