Roman Lutz

@SeanMcCarren > In Exponentiated gradient, how can we know if y is binary or continuous? If it is continuous, we should not apply the encoding. Can we do the same...

> Exponentiated gradient testcase is dependent on 0/1 predictions in Moments, I don't know how to fix this. For now, I'll leave the prediction as 0/1s. You can adjust those...

This is awesome, thanks for the summary! First of all, I would create separate issues for the various algorithms once you've decided which to implement. Starting from the metrics sounds...

FWIW I am always in favor of providing as many metrics as possible, explaining their benefits and drawbacks (e.g., through a user guide or example), and then letting users choose...

FWIW I didn't mean to make you run entire studies against benchmarks. I merely meant that the most meaningful metrics are those that capture downstream harms so we don't need...

I have still not had time to read through it in detail, but it does seem like it would require people to manually translate pages. I'm unsure what will happen...

@candalfigomoro You can actually compute this with `MetricFrame`! ``` >>> mf = MetricFrame(metrics=mean_squared_error, y_true=[1,0,1,0], y_pred=[0,1,1,1], sensitive_features=['a','a', 'b', 'b']) >>> mf.by_group sensitive_feature_0 a 1.0 b 0.5 Name: mean_squared_error, dtype: object >>>...

We've had this discussion a while ago, although I'm having trouble tracking down where exactly. What I remember from that time: - CHANGES.md is meant for ALL changes to the...

I've started taking a stab at this. My upcoming PR moves the previously called CHANGES.md to the website (it's a lot harder than it sounds! 🤣 ). From there we...

Closing this issue since there's no response. Feel free to reopen if you've tried the suggestion. Please also provide details on your OS, Python version, and relevant Python package versions...