Title :
Adaptive integratable hardware realization of analog neural networks for nonlinear system
Author :
Zhan Su;Bogdan M. Wilamowski;Ruixin Wang;Fa Foster Dai
Author_Institution :
Department of Electrical &
fDate :
7/1/2015 12:00:00 AM
Abstract :
This paper presents the adaptive analog hardware implementation of a MLP (multilayer perceptron architecture) ANN (artificial neural networks) for online nonlinear system operation. Neurons are implemented by bipolar differential pairs with tangent hyperbolic activation function. A bipolar current multiplier and a linearized differential amplifier are proposed for storing and adjusting the weights for ANN where its input current can be adjusted or reprogrammed by outside digital controllers. Compared with other hardware-based MLP implementations, it provides a better cost efficient ANN platform that can be fully integrated on chip while keep the network with high performance with high frequency requirements. Such an ANN platform can be adapted for differential applications on control or nonlinear model systems without changing the architecture.
Keywords :
"Artificial neural networks","Hardware","Biological neural networks","Computer architecture","Neurons","Transistors","Topology"
Conference_Titel :
Industrial Informatics (INDIN), 2015 IEEE 13th International Conference on
Electronic_ISBN :
2378-363X
DOI :
10.1109/INDIN.2015.7281788