Implementação de redes neurais Perceptron e Adaline em ambiente LabVIEW™

Abstract

Artificial Neural Networks (ANN) are specialist sistems which have structures inspired in biological neurons, which due their capacity to process complex data and constant learning are widely used for solving problems. This capacity for problem solving has several areas of operation such as finances, economics, factory plant control, and pattern recognition. With the technological evolution the programing of an ANN can be made in different microcontrolled plataforms. A very used technique for prototyping, test, and for studies is the virtual instrumentation. Virtual Instrumentation is defined in the use of softwares that simulates the assembly of a circuit or a chain of logical operations without the need of the physical assembly. This technique allows to saving time and resources, whilst still have the advantage of being edited, copied, simulated without any kind of physical resource other than the computer and the program for the virtual instrumentation. The study of this work intent to implement and compare Perceptron and Adaline ANNs in the softwares MATLAB® and LabVIEW™, as a way to observing the construction process and the form of presentation in both programs. For such results to be achieved, databases witch known outputs and inputs ware used. Both samples were tested in both programs. With the results of the tests, they were compared and it can be observed that the MATLAB® environment will be more effective in processing time, however, the LabVIEW™ program presents the expected results and still provides an external data acquisition platform more practical and versatile.