An analogue ANN for classification of alcohol
Y.C. Leung, D.H.F. Yip, W.W.H. Yu · 2002
The paper presents the practical implementation of a neural network for alcohol classification using an array of operational amplifiers (OAs). Backpropagation training is used to obtain the weight matrix. There are output voltage constraints for operational amplifiers. Applying output level constraints to the nodes at the simulation or training process produces the correct weight matrix for an OA-based neural network classifier. The paper shows that node output level constraints are important for designing a neural network using an operational amplifier. Because operational amplifiers are low-cost off-the-shelf components and the implementation is relatively easy, designing commercial neural network classifiers using OAs could be an attractive alternative to neural network ICs.