Zayd Hammoudeh

One of the primary reasons you see the same list of opponents/proponents for different test examples is that the mislabeled/atypical training instance have gradient magnitudes that are significantly larger than...

I agree with @alexandertornede. ACML is growing in prominence with increasingly good research published there and prominent PCs (e.g., Masashi Sugiyama). ECML may also be worthwhile if we add ACML.

@henrygouk Agree. ECML is more prestigious and established than ACML.

After submitting this case, I noticed that `scale` and `damp` have different values listed as default than in `get_default_config`. Both are `None` in the code but in the README it...

@iamgroot42 added a `NaN` check in his fork to detect when there is numerical instability. Those are probably worth rolling into a pull request (either this one or a separate...