You might want to have a look at stochastic gradient descent, where the error is propagated back after every sample (or mini batch).

Thanks again Davide, :)

Hi Ahsan,

I usually use the standard backpropagation with the gradient descent

algorithm in training, that means the errors

are passed back down the network and the

weights between the neurons are adjusted in the direction

of the negative gradient of the performance function. The literature describing the details of this commonly used learning algorithm is very extensive

(see for example :

Demuth H, Beale M. Neural network toolbox users guide. Math-

Works, Inc.; 2003.

Veelenturf LPJ. Analysis and applications of artificial neural

networks. Prentice-Hall Intl (UK) Ltd; 1995. p. 70–6.

I hope that this helps.

Good luck with your work!

Luiza

In my experience epoch based training is very coarse and, in addition to the problems you have identified, it does not permit fine tuning of weights on an example by example basis. Like the two previous respondents I would suggest you adjust the weights after each presentation.