How to re-train a neural network

3 views (last 30 days)
Israel Yaish
Israel Yaish on 29 Apr 2021
Answered: Aditya Patil on 10 May 2021
How to re-train (continue training) a neural network using the following implementaion:
net = trainNetwork(XTrain,YTrain,layers,options);

Answers (1)

Aditya Patil
Aditya Patil on 10 May 2021
Check the documentation on how to resume training from checkpoint.
In short, save the model to a mat file using the CheckpointPath option, and load the mat file later. Then you can resume the training my passing the layers of the checkpoint during training, as follows
trainNetwork(XTrain,YTrain,net.Layers,options)
Note that if you reuse the same options, you might see that the training diverges, as the training might have reduced the learning rate. Hence, use appropriate options as required.

Categories

Find more on Image Data Workflows in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!