Open
Description
We have a hypothesis that line search for hyperparameter (coarse_parameter_tuning
) phase is already doing such a good job, that the identified feasibility interval does not contain a solution that could improve the performance further (module random variation).
To investigate this hypothesis, we need to look at the logs of hyperparameter optimization path for some real-world scenarios.
- Is additional logging required?
- Can I run the jobs in the cloud myself and look at the logs?
- Summarize results in a google sheet.