Add customized objective (loss function) and eval metric in XGBoost

[exalate-issue-sync[bot]]

Currently we can only specify loss function through distribution and evaluation metric through stopping_metric. But original XGBoost package supports customized loss function and eval metric, see [here|https://github.com/dmlc/xgboost/blob/master/demo/guide-python/custom_objective.py]. It would be great if H2O XGBoost could support the feature.

[exalate-issue-sync[bot]]

Michal Kurka commented: This will be partially implemented in [https://h2oai.atlassian.net/browse/PUBDEV-8868|https://h2oai.atlassian.net/browse/PUBDEV-8868|smart-link]

[h2o-ops]

JIRA Issue Migration Info

Jira Issue: PUBDEV-6579 Assignee: New H2O Bugs Reporter: Zhifeng Wang State: Open Fix Version: N/A Attachments: N/A Development PRs: N/A

[wendycwong]

Support ticket: https://support.h2o.ai/a/tickets/104467

[valenad1]

@wendycwong This issue is slightly different from the requirement in supported ticket. Its about integration with xgboost library, the H2O scoring is different.

I will create another issue for the support ticket requirement.

#15967