Implementation of Hyperbolic Tangent Activation Function in VLSI
I P. Motcha, Mary Rathi Pushpa, K. Manimala · 2014
This work presents the implementation of Artificial Neural Network (ANN) chip, which can be designed to implement certain functions. Usually designing of neural networks is done using software tools in the computer system. The neural networks designed off-line are fixed and are very slow especially for portable real time applications. In order to overcome this disadvantage, neural network is to be implemented on-chip with the neural network. Efficient implementation of the activation function is important in the hardware design of artificial neural networks. In this project, a simple and efficient approximation for digital hardware implementation of the hyperbolic tangent function is presented. The proposed technique employs a hybrid method which is the combination of linear method and bit level mapping and found to perform better than other conventional method.