How to save the criterion at which neural net training was finished?

2 次查看（过去 30 天）
Alexandra Aulova
Alexandra Aulova 2021-2-2
评论： Alexandra Aulova 2021-4-7
Hello!
I am training a fitnet network with trainbr. I run training of different topologies in cycles for nets with different number of neurons and when trained save customly calculated performance parameters for further optimization. The question is: how can I save the criterion on which training process was exited? In whatever form will be ok. This will help me analyze the results, as i will be able to adjust training parameters, now i have them all at default values as the first iteration. Particularly I am interested if the training was finished based on number of epochs or other performance parameters.
thank you in advance for your answers!
Alexandra
  1 个评论
Alexandra Aulova
Alexandra Aulova 2021-2-12
Dear all,
Is it possible at all? Maybe @Greg Heath could help?...
I will be very grateful for this inforamtion since I am looking for it for some time already...

请先登录，再进行评论。

请先登录，再回答此问题。

回答（1 个）

Swetha Polemoni
Swetha Polemoni 2021-3-30
Hi
With trainbr, training occurs according to trainbr training parameters. You might find the following documentation useful.
https://www.mathworks.com/help/deeplearning/ref/trainbr.html
  1 个评论
Alexandra Aulova
Alexandra Aulova 2021-4-7
Thank you for the link. This unfortunately does not help, since I am looking for the option to save the training exsiting criterion, either number of ephoch or mu parameter or MSE target for example.

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

标签

Community Treasure Hunt

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

Start Hunting!

Translated by