How to return a final training loss value?

Question

Sho Wright 2021-4-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/793952-how-to-return-a-final-training-loss-value

回答： Lokesh 2023-10-13

I want to evaluate the results of my network at various settings (changing batch size, maximum epochs, dropout layer rates,etc).

I've set my 'verbose' to true to extract data as to obtain mini-batch and validation RMSE/loss data per epoch, and I want to compare my final validation loss against my final training loss, as from what I know these can be compared to see if your network has resulted in underfitting, overfitting or optimally.

I have used 'info' to return my final results but the training loss is stated as a single column

Is there any way to get a final training loss, the same way I have a final validation loss and RMSE?

Thank you

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

Lokesh 2023-10-13

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/793952-how-to-return-a-final-training-loss-value#answer_1332484

Hi Wright,

I understand that you want to obtain the value of the final training loss. However, ‘info.TrainingLoss’ consists of a column of values rather than a single value.

The ‘info.TrainingLoss’ variable contains a column of values, where each element corresponds to the loss for a particular iteration. For example, ‘info.TrainingLoss(1)’ corresponds to the training loss for the first iteration, ‘info.TrainingLoss(2)’ corresponds to the training loss for the second iteration, and so on. By accessing ‘info.TrainingLoss(end)’, you can retrieve the value of the final training loss.

I hope you find this helpful.

Best Regards,

Lokesh