How to retrain a CNN/LSTM using new recorded data?

Question

CHRISTOPHER MILLAR 2021-10-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1468911-how-to-retrain-a-cnn-lstm-using-new-recorded-data

回答： CHRISTOPHER MILLAR 2021-10-18

I have trainined a hybrid CNN/LSTM network to classification sequential data, in offline experiements it works very well.

In a live scenario it does not perform as well so i wanted to take the sequential sequences recorded during the live experiments and then combine this with the original dataset and retrain the model with the combined dataset.

Once I combined the old dataset with the new data i tried

newNet = trainNetwork(oldNet,XTrain,YTrain); & newNet = trainNetwork(oldNet,XTrain,YTrain,layers,options);

This gives me an error of too many input arguments

Is this possible?

Regards

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Mahesh Taparia 2021-10-15

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1468911-how-to-retrain-a-cnn-lstm-using-new-recorded-data#answer_809586

Hello,

You can refer this documentation of training from pretrained weights. You can save the trained network from the old data and load it while adding the new data.

Hope it will help!