Michaël Defferrard
Michaël Defferrard
Let's think a bit more about the name. If we'll group them, you can name the module `graphs/learned.py`.
Good to hear from you @kalofolias! Thanks for the tutorial. Do you like `LearnedFromSmoothSignals` as a name? Any other idea? Are there settings where we'd prefer Dong et al. over...
Thanks for your patience. @kalofolias, you should have received an invitation. Upon accepting, you should be able to push your changes to this branch. Thanks for the tutorial!
Is @kalofolias actively using this? Can we push his improvements and tutorial while we have it in our mind? Thanks :)
@nperraud: why wasn't this bug caught by the unit tests? Can we add some to avoid further issues?
I'm not sure whether that's in scope. But IIRC they use GSP / spectral graphs concept. Right? Should be pretty close to our `features.compute_spectrogram()`. So it could go in that...
Thanks! > It is essentially the characteristic function of the tig. That's good. Waiting on @nperraud to weight in. Then we can see how to integrate it properly.
I too believe it should be quite similar. `compute_spectrogram()` was developed by @nperraud. AFAIK it has only been published in [his thesis](https://perraudin.info/publications/perraudin-note-019.pdf) (chapter 3). ![20210320_035426](https://user-images.githubusercontent.com/6806065/111857166-fca3e600-892f-11eb-8002-a482750a897b.png)
Shall we consolidate `compute_spectrogram()` and GraphWave in the PyGSP?
Indeed, people are confused (e.g., #67 and #68). But making a new release won't solve the issue in the long run. The development version will always be ahead of the...