Keunwoo Choi

typos .. `retunrs`

- [ ] random gain (control the multiplication factor) - [ ] random volume (control the output.. amplitude? energy?)

- [ ] more examples - [ ] pre-structured models?

Yes, you’re right. > On Aug 5, 2021, at 06:23, Bagus Tris Atmaja ***@***.***> wrote: > >  > Is that tthe following layer is the replacement? > composed_melgram_layer =...

Hi 😊 thanks for the issue and the context. Maybe one thing we can do is to wrap the numpy fft function and selectively use it based on the hardware...

@JPery Sounds like not a bad idea. To maximize its utility, we'd want this to work i) automatically when it's on cpu ii) without any complex configuration iii) when there...

@munael I also hope we have this functionalities. That said, in scikit-learn, `groups` parameter ignored and maybe it's the same here.

On MacPro 2014 (no multi processing) with 10 wave files. CNMF took 1555 seconds and Spectral clustering: 211 seconds. Posting for someone who might be also interested, performance was not...

AND, for the tuner, shouldn't I pass an EarlyStopping callback? I suspect if it is stopping the training after a few epochs but the tuner doesn't know about it, hence...