A Novel Method for Scalable VLSI Implementation of Hyperbolic Tangent Function

Hyperbolic tangent and Sigmoid functions are used as non-linear activation units in the artificial deep neural networks. Since, these networks computationally expensive, customized accelerators designed for achieving required performance at lower cost power. The function MAC key building blocks of A low complexity accurate hardware implementation is to meet area targets such network accelerators. Moreover, a scalable recent studies show that DNNs may use different precision layers. This paper presents novel method based on trigonometric expansion properties hyperbolic which can be easily tuned accuracy requirements.