What happened to the accuracy before and after 75th? Why before 75th, its unstable and suddenly stepped up after that? I am using RESNET and CIFAR-10 dataset
.
$\begingroup$
$\endgroup$
1
-
$\begingroup$ Did you change the learning rate after 75th epoch? Which optimization algorithm do you use? How much data do you have? What's the batch size? What's the beginning learning rate? Which loss are you using? How did the loss change after 75th epoch? All these things are necessary to be able to answer your question? $\endgroup$– Antonio JurićCommented Mar 14, 2019 at 8:36
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
2
I think that it is a convergence of the gradient descent happened. While training, we are looking for correct weights and that is why we see these fluctuations on the picture. When we are close to a global minimum (found good weights), the model starting to converge. That is what happened after the 75th epoch in this case.
By the way, you have an overfitting
-
$\begingroup$ how do we know its overfitting? and how to avoid it? $\endgroup$– ErmeneCommented Mar 13, 2019 at 8:19
-
$\begingroup$ The training accuracy is very close to 1 and the validation accuracy is lower than training. That is a sign of overfitting. To avoid it you can try to collect more training data, use regularization, dropout. $\endgroup$ Commented Mar 13, 2019 at 9:18