Roman Lutz

Add unique output block references for literalinclude

10

It's sometimes useful to reference code block output in addition to produced images (using `figure`) or code lines (using `literalinclude`), so this change adds comments to the produced RST file...

Referencing output from cells

3

I can reference code from a notebook using ``` .. literalinclude:: ../auto_examples/plot_myscript.py :start-after: [some-tag] :end-before: [another-tag] ``` and the resulting images using ``` .. figure:: ../auto_examples/images/sphx_glr_plot_myscript_001.png :target: ../auto_examples/plot_myscript.html ``` However,...

Simplify route generation logic for testing with flights

12

## Description So far, we used the string `"noFlights"` to avoid providing an empty string that the route generation can't handle which is not ideal. Instead, this PR removes the...

Model overview: add box plot in the regression case

3

## Description Box plots are so far only there in the binary classification case, but this PR adds a box plot for regression, too! It supports error, predicted y and...

Add user guide for fairlearn.postprocessing.ThresholdOptimizer

10

This PR finally adds a user guide for `ThresholdOptimizer` which explains some of the pitfalls that users should be aware of when using this. If this reads like you shouldn't...

Add user guide section on persisting metrics using pickle

3

#### Describe the issue linked to the documentation Persisting metrics and loading them back into memory is often useful. We don't currently mention this anywhere but `fairlearn.metrics.MetricFrame` can be pickled...

Since our Azure DevOps tenant is orphaned and we can't seem to get it back we have to consider migrating to another tenant or completely off of ADO towards GitHub...

Nightly requirements-fixed builds are failing

2

The title says it all. We need to look into that and get it fixed.

DOC: styleguide for referring to groups of people

4

I've noticed we that we have a few occurrences of terms like "blacks", "whites", "females", "males", and perhaps others in our documentation. This made me aware that I don't know...

Expectations for predict, predict_proba, etc.

6

Since fairlearn uses randomized classifiers we need to set expectations for what `predict`, `predict_proba` etc. return.