Open AccessAccuracy controlled iterative method for efficient sigmoid function approximation
Author(s) -
Chiluveru S.R.,
Tripathy M.,
Mohapatra B.
Publication year - 2020
Publication title -
electronics letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.375
H-Index - 146
ISSN - 1350-911X
DOI - 10.1049/el.2020.0854
Subject(s) - sigmoid function , activation function , function (biology) , linearity , piecewise linear function , computer science , piecewise , algorithm , function approximation , iterative method , approximation error , nonlinear system , artificial neural network , mathematics , mathematical optimization , artificial intelligence , mathematical analysis , electronic engineering , engineering , evolutionary biology , biology , physics , quantum mechanics
The sigmoid activation function plays an essential role in implementing neural networks in hardware. However, due to a high degree of non‐linearity, the hardware implementation of the activation function itself is a daunting job. In this Letter, the authors proposed an accuracy controlled iterative algorithm for the piecewise linear approximation of the sigmoid function. The proposed algorithm approximates sigmoid activation with an optimised number of linear functions for a given error. The accuracy and hardware complexity of the resulting approximation function are compared with the state‐of‐the‐art literature.