Would it be possible to update a trained network with an updated training set for active learning purposes without reatraining it from scratch?

Hi, I would like to make some tests of active learning with matlab neural networks. For now what I am doing is:

Get a training set Train the network from scratch Calculate the error on another dataset Update the training set with new point(s).

Is there any way of retraining the network from a previous network, in order to update the weights and biases. I guess just updating and not retraining from scratch would lead to a faster training.

Regards,

Florian