The world's cleanest AutoML library ✨ - Do hyperparameter tuning with the right pipeline abstractions to write clean deep learning production pipelines. Let your pipeline steps have hyperparameter spaces. Design steps in your pipeline like components. Compatible with Scikit-Learn, TensorFlow, and most other libraries, frameworks and MLOps environments.
Describe the bug
In the Orthogonal (SVD) TPE (see #464), we need discrete distributions to be made continuous. This is ok for some distributions (e.g.: quantized distributions, such as RandInt), however for some other distributions, there is no linear (growing) dependency between the values
To Reproduce
E.g.: in a Choice, there is no transitive relation between the item 0, 1, and 2 of the Choice. This doesn't happend in a RandInt.
Expected behavior
The Choice hyperparam to be encoded as a one-hot. And its rounding (as per #466) after undoing the SVD would need to pick the argmax of this one-hot.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs in the next 180 days. Thank you for your contributions.
stale[bot]
commented
1 year ago
Describe the bug In the Orthogonal (SVD) TPE (see #464), we need discrete distributions to be made continuous. This is ok for some distributions (e.g.: quantized distributions, such as RandInt), however for some other distributions, there is no linear (growing) dependency between the values
To Reproduce E.g.: in a Choice, there is no transitive relation between the item 0, 1, and 2 of the Choice. This doesn't happend in a RandInt.
Expected behavior The Choice hyperparam to be encoded as a one-hot. And its rounding (as per #466) after undoing the SVD would need to pick the argmax of this one-hot.
Additional context
464