Application of neural networks for on-line calculations
Valeri Mladenov, Evangelos Zirintsis, Christos Pavlatos, Vasiliki Vita, Lambros Ekonomou · 2009
Modern electronic devices where a touch screen is implemented, require the use of a more straightforward calculation system, the on-line calculation system. The formula input in such a system is in an image form. The requirements of the data-set and how these symbols are preprocessed are described respectively. In order to process the image and perform the correct calculations, the images are used as an input to Neural Networks. Two different Neural Networks are trained, the first is an autoassociative memory Neural Network (NN1) and the second a multi-layer feed-forward back-propagation Neural Network (NN2). Finally, both Neural Networks are tested and the total performance of the application is illustrated.