Gilles Louppe

For ease of reuse and consistency, what are your opinions about defining styles somewhere globally and then import them in the examples? I am thinking about something like `plt.xkcd()` but...

or we might check why that many points are reevaluated.

Hi Matthew! Yes, I believe this would be worth fixing. Although more importantly, I think the project needs to find new maintainers and release managers :-)

+1 for that extra argument. Just hit the same problem myself, writing a closure is not always the nicest thing for code organization...

Hi! Thanks for your interest. Are you suggesting to convert the whole thesis into a giant IPython notebook? This would be great but I am afraid this would also take...

Okay then, feel free to reuse them to write notebooks :)

Thanks Lukas! I'll try to see what's wrong... (nothing was changed). For the demo you wanted to show, maybe https://github.com/glouppe/notebooks/blob/master/Fair%20classifier%20with%20adversarial%20networks.ipynb would work? (although I do remember bugs with newer versions...

I just did a quick benchmark using the text files as input features. 100 extra-trees: ``` mfcc = 0.92995 (0.03840) ceps = 0.93973 (0.03060) specs = 0.81603 (0.02211) ``` Both...

On the rastamat page, they also talk about "deltas" computed over cepstras. ~~Might be worth adding them as well.~~ Update: ``` deltas = 0.53022 (0.09145) ```

Using a sampling rate of 4kHz instead of 2kHz to generate cepstra and spectra reduces the number of features rhoughly by a factor of 2 (9_98 features for cepstra, 13_98...