CODEWITHSUNDEEP
mathmachine-learning
Imran
Imran

Reputation: 99

Retrain of a neural network

I have already trained a NN with training data set [X,Y]. Lets consider this model is called M. Now, I have a new data set [Xnew,Ynew]. I would like to Update my model M with data set [Xnew,Ynew] without retraining it from scratch.

My question is: is it possible to update the model this way or I must retrain this? If it is possible to update, how?

Upvotes: 1

Views: 405

Answers (1)

Tamás Zahola
Tamás Zahola

Reputation: 9321

Yes, if you’re doing e.g. backpropagation. You can take the weights trained on the first data set as the starting point for the training on the second data set, instead of starting with random weights.

Whether this is a sensible thing to do or not is of course highly depends on your data sets and the network topology.

Upvotes: 2

Related Questions