fjarri/reikna

Add a link to the discussion forum to README.md

flavianh opened this issue · 4 comments

Before anything, I'd like to say that you did a very good job on this library and its documentation!

I'd like to compute a spectrogram in a similar way as with mlab. Basically it consists in a FFT taken at different times in an audio signal.
I am not yet very familiar with the notions of transformations and computations, and I'd like your advice on getting things right as to where I should put a rolling window in the computation tree. Let's say I use your FFT function, what should the tree look like if I have a computation named rolling_window?

It could also be interesting to know your thoughts on how to apply this to a more general kernel. Would you use a rolling window as well, or stack your signal vector in a matrix (stacking each window on top of the other)?

Thank you, I tried to explain the transformation/computation model in the docs, but apparently did not do a very good job of it :)

I've compiled an example of a specgram analogue: examples/demo_specgram.py in this repo. If you want an exercise, read the initial docstring where I outlined the logic and try to implement the computation yourself.

(For the future, questions like this are better suited for the discussion forum (since it's not really a bug))

Sorry, I missed the discussion forum on your doc, I usually find those on the README.md file in the repo, I guess that could be my contribution for an issue ;). The example looks great, I had missed it too, thanks!

You didn't miss it, I just added it in response to this issue :)

P.S. Speaking of the example, the padding support was missing. Fixed now.