Use vectorization and numba jitting for optimization

[kp992]

This PR optimizes the runtime of black_litterman lecture by using vectorization and numba jitting.

Fixes #183

[jstac]

Thanks @kp992 , this code looks much nicer. @mmcky could you please check the outputs?

[kp992]

I see that the build time is too high still. Locally its working fast for me. May be some issues with the ipywidgets? The slider is not working https://python-advanced.quantecon.org/black_litterman.html. See the SS

[kp992]

The slider is working locally for me but it doesn't on live pages because the html pages are static in nature.

[kp992]

I think the high timing issue is not consistent in this lecture. I added a debug commit a9c061f#commitcomment-149458134 and it just took 42s (https://674294b5a50f0309d83896e3--wonderful-lalande-528d1c.netlify.app/status).
After reverting the same commit, its ~7200s (https://6742b3a5c174582b302b578c--wonderful-lalande-528d1c.netlify.app/status).

This looks strange to me.

[mmcky]

thanks @kp992.

We should remove the sliders as they won't work on the static site representation. Perhaps we can put the slider version in a collapsed code-block (and tag no-execute) and add a note block that says if you are running this lecture locally you can adjust the parameters using the piython slider.

Then we can see if the intermittent issue persists?

[mmcky]

@kp992 I have just noticed that matsuyama also has a very long run time of 7000s. It may be worth checking if they both have something similar (like the widgets) that is hanging the execution?

matsuyama | 2024-11-04 08:33 | cache | 7206.91 | ✅ -- | -- | -- | -- | --

matsuyama
2024-11-04 08:33
cache
7206.91
✅

[kp992]

Thanks @mmcky. Yes, I also believe that ipywidgets is the one hanging the build.

[mmcky]

thanks @kp992 this is looking great.