FPGA based implementation of sigmoid function using different approaches [Farkli yaklasimlar kullanarak sigmoid fonksiyonunun FPGA tabanli gerceklenmesi]

Abstract

In this study, FPGA based implementation of sigmoid activation function, which is frequently used in Artificial Neural Networks (ANNs) applications, is performed by using two different approaches. Designs have been coded using VHDL. For the representation of real numbers in digital systems, a 32-bit single-precision floating-point number system is used. Basic arithmetic units such as addition, multiplication and division have been created for these number systems. In this paper, Altera Cyclone V FPGA integrated circuit has been used for designing feedforward artificial neural network to solve XOR problem. As a result of the study, it is seen that the first approach, which is the piecewise linear approach, produces results for XOR problem with zero error. © 2020 IEEE.