Bokeh application in Jupyter with Docusaurus output

Wanted to highlight a tool that is beginning to see some use.

It is a part of Bean Machine (a probabilistic programming language [PPL]) that is using Bokeh, BokehJS, and custom JavaScript callbacks for user interaction. The tool highlighted in the link shows the Kernel Density Estimate (KDE) of the posterior distribution of a Bean Machine model. Users can interact with the bandwidth used to calculate the KDE, and they can modify the Highest Density Interval (HDI) to see how changing these parameters affect the results.

The tool is directly viewable in a Jupyter notebook, and it can be converted to an MDX file that Docusaurus can render. The above link is the Docusaurus rendering of the notebook and the link below

is close to a one-to-one representation of the notebook to the Docusaurus page. The only difference being that the Docusaurus page adds two links at the top of the notebook; one for colab and one for the GitHub page of the tutorial.

There was a lot of tools used to make this simple Bokeh application, but I think it highlights some of Bokeh’s capabilities. There are other tools in the works, but this is the first to be viewable by everyone. I hope y’all enjoy.

1 Like

This is super cool thank you far sharing! We’d love to tweet out some of these if that’s ok?

1 Like

Certainly. Feel free to share on any platform you’d like