Update GLM Demo to use predict_proba
Open
- Dominant language
- C++
- Stars
- 468
- Forks
- 96
- PR merge metrics
- No merged PRs in 30d
Description
In the H2O4GPU_GLM demo, the predictions are calculated by calling predict on the GLM model. This returns the class predictions, not the probability predictions, and causes the calculated logloss to be Inf.
Change the demo to return the predicted probability instead of predicted class.
Contributor guide
Assessment
This issue has not been assessed yet.