In principle, 'training' a neural network, be it recursive or feed-forward, is simply optimisation of the network parametrs with respect to a particular cost function. Viewed in this light, back-propagation is actually an instance of the stochastic gradient descent algorithm which has this special structure of 'back propagating errors' due to application of the chain rule. BPTT is SGD for recursive networks. So, yes, you can use SGD, but perhaps a better idea would be to employ some other optimisation algorithm instead.

Prashant,

I would suggest reading the paper by Paul Werbos "Backpropagation Through time: What it does and how to do it". It is an old paper but still the best reference on the subject (and has segments of a Fortran program which makes it easier to implement).

Hope this helps

In principle, 'training' a neural network, be it recursive or feed-forward, is simply optimisation of the network parametrs with respect to a particular cost function. Viewed in this light, back-propagation is actually an instance of the stochastic gradient descent algorithm which has this special structure of 'back propagating errors' due to application of the chain rule. BPTT is SGD for recursive networks. So, yes, you can use SGD, but perhaps a better idea would be to employ some other optimisation algorithm instead.

Hi, Since the basic back-propagation (BP) learning algorithm is too slow for most practical applications, there have been extensive research efforts to accelerate its convergence. Different gradient-descent variations of the backpropagation algorithm have been considered like Levenberg-Marquardt (LM), Scaled Conjugate Gradient (SCG), One Step Secant (OSS), and Quasi-Newton BFGS. So, it's rare to find someone using the original BP algorithm even with feedforward networks. Yes, it is possible to apply any one of these variations of BP to recurrent neural networks. I also recommend trying reservoir computing algorithms like: echo state machine, backpropagation decorrelation, and liquid state machine. BPTT also can be utilized with some recurrent models like Elman network. Good luck.

Try with NARX network and read the papers on this algorithm. It will help you to get the network with BP based RNN