0

I use Siamese Network to train my own dataset.I visualized the training results using Tensorboard.I'm a beginner and I'm not very good at judging if it's an overfit.Please help to see, thank you! Figure 1 and 2 are Train. Figure 3 and 4 are Validation. Figure 1 This is the accuracy of the training set

Figure 2 Train Loss Figure 3 The accuracy of the Validation set Figure 4 Validation Loss

Wyuer
  • 11
  • 4
  • 1
    Those pictures without legend or axis labels won't make sense to anyone. About overfitting and underfitting, please google https://machinelearningmastery.com/overfitting-and-underfitting-with-machine-learning-algorithms/ – Innat Mar 22 '21 at 02:09
  • hi.I looked at the link you shared.But I'm a little confused about how to judge.How do I tell if I'm overfitting? Do you just look at the predictions?Looking forward to your reply. Thank you – Wyuer Mar 22 '21 at 02:49
  • What makes you confused? What's your current understanding about overfitting or underfitting? – Innat Mar 22 '21 at 03:21
  • My understanding is that an overfit is a function that is not fitted, and when I look at Figure 3, it looks a bit like an overfit. – Wyuer Mar 22 '21 at 03:32
  • I see. Please see this QnA, https://stackoverflow.com/questions/52009816/how-to-know-if-underfitting-or-overfitting-is-occuring – Innat Mar 22 '21 at 03:35
  • 1
    I understand after reading your share, thank you so much. – Wyuer Mar 22 '21 at 07:15

1 Answers1

0

Your network does NOT appear to be over fitting. A classic sign of over fitting is when your training loss continues to decrease while your validation loss begins and continues rising

Gerry P
  • 7,662
  • 3
  • 10
  • 20