Skip to content
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

experiment result explanation in tutorial #65

Open
liangchen1ceeee opened this issue Feb 4, 2021 · 2 comments
Open

experiment result explanation in tutorial #65

liangchen1ceeee opened this issue Feb 4, 2021 · 2 comments

Comments

@liangchen1ceeee
Copy link

Hi, I'm looking at the shape constraints tutorial, and the results for GBDT and DNN are listed in the tutorial as follows:

  • GBT Validation AUC: 0.7248634099960327
  • GBT Test AUC: 0.6980501413345337
  • DNN Validation AUC: 0.7518489956855774
  • DNN Testing AUC: 0.745200514793396

After the experiment results, the tutorial comments Note that even though the validation metric is better than the tree solution, the testing metric is much worse.

I don't understand where this comment comes from, since DNN outperforms GBT in both validation AUC and testing AUC.

@mmilanifard
Copy link
Collaborator

Thanks for pointing out the issue. The text matched the results at the time of writing the colab, but due to changes in the training for GBT or DNN models the results have moved around a bit. The primary reason for this instability is that we cannot train the models until convergence as the tutorials are auto generated from the colabs and need to run in a few minutes.

We are updating the colab to use less data and a more aggressive learning rate to get closer to convergence, re-optimizing the hyper parameters along the way. These changes will be reflected in the upcoming release.

@liangchen1ceeee
Copy link
Author

Thanks. Look forward to your new results!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants