# How can recurrent neural networks be used for sequence classification?

RNN can be used for prediction, or sequence to sequence mapping. But how can RNN be used for classification? I mean, we give a whole sequence one label.

• I would try to predict the next term using the previous. I would then prepend that network to a radial basis function (of 3*classlabelcount or so gaussian pdf neurons) whose inputs are the prediction errors of the series and whose outputs are the class labels. I would have a decent number of neurons in the input layer - enough to account for how many elements of the sequence are required to classify. This is just my personal approach. Dec 17 '14 at 3:38