-
Notifications
You must be signed in to change notification settings - Fork 4
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
why the outputs of the RethinkNet is all close to zero? #1
Comments
Probably not "all" are close to zero? |
Thanks for your reply. The outputs of all labels are below 0.1 so i wander if the wrong model was built. The model is shown as follows:
` |
Have you train the model and have the loss converged? |
I trained it 300 epochs, and the loss converges to 0.0746 |
Which cost function are you training the model with? |
The cost function is binary crossentropy. So i dont understand why the outputs have no labels close to 1. |
For training the RethinkNet, it is in fact training on a weighted binary crossentropy loss. You can check the implementation here. If using the binary crossentropy alone without changing the weight. |
Thanks. |
I trained the net on bibtex dataset.
The loss function is binary crossentropy.
Thank you.
The text was updated successfully, but these errors were encountered: