A partially recurrent architecture applied to classification problems
M.B. de Martino · 2002
Neural networks are a promising tool for artificial intelligence applications, which mostly can use some kind of classification in their solution. Therefore, we discuss the necessary requirements for applying neural networks on classification problems and present a new partially recurrent architecture based on Jordan and Elman's models. We then select and use the "backpropagation through time" algorithm on the proposed architecture and test it in an example given by Telfer.