Open AccessSOFTSWISH NEURAL NETWORK APPROXIMATION WITH ZUI-CUI MODULUS
Author(s) -
Hawraa Abbas Almurieb,
Zainab Abdulmunim SHARBA
Publication year - 2021
Language(s) - English
Resource type - Conference proceedings
DOI - 10.47832/minarcongress3-7
Subject(s) - artificial neural network , multivariate statistics , smoothness , function (biology) , function approximation , key (lock) , modulus , activation function , modulus of continuity , mathematics , computer science , algorithm , artificial intelligence , mathematical analysis , machine learning , type (biology) , geometry , geology , paleontology , computer security , evolutionary biology , biology
Until today, many formulas of neural networks are defined to be used for function approximation, they vary with respect to the weights, activation functions and other standards. Moreover, researchers have studied the approximation of different spaces of functions. In this paper, we approximate functions from multivariate spaces with a neural network with a new defined form of Swish function, named SoftSwish. Also, multivariate Zou-Cui modulus is introduced to express the degree of approximation by our Swish neural network that we call “SoftSwish Neural Network”.. Key words: Approximation, Neural Network, Swish, Modulus of Smoothness.