David Holzmüller

Wrong number of tasks displayed for AutoML All Classification

1

**Description** The ["Task Collection" tab for the "AutoML Benchmark All Classification" benchmark](https://www.openml.org/search?type=benchmark&sort=tasks_included&study_type=task&id=271) (id=271) says "66 tasks", while the "Tasks" tab lists 71 tasks. The corresponding paper and the Python API...

Implement temperature scaling for (multi-class) calibration

32

### Describe the workflow you want to enable It would be great to have temperature scaling available as a post-hoc calibration method for binary and multi-class classifiers, for example in...

CatBoost overrides scale and bias

3

Problem: CatBoost appears to override the values I provide in `set_scale_and_bias()`, even if I set `boost_from_average=False`. In the latter case, it sets scale and bias to $(1.0, 0.0)$. Minimal working...

Limit Cramer's V analysis to target column

15

### Problem Description I would like to get a quick idea of which features are important for predicting the target. For this, it would be nice if the TableReport had...

Implemented adaptive squashing

13

WIP