Interactive graphs + making hyperparameters function variables to show various effects

[mariemayadi]

One ask from the meeting notes state: hyper parameter effect.

[mariemayadi]

Team meeting brainstorming:

• dashboard?

[mariemayadi]

To be discussed: output into data structure of current code

[mariemayadi]

As a start @kwomackcodes will explore visualization/dashboarding.

[mariemayadi]

Research Ideas:

PYTHON From experience, for visualization, in python the go to these days is DASH plotly (it's basically R Shiny but for python). The advantage there is potentially having less of the overhead mentioned by @Benjamin Weiss Livingston around needing to convert output of existing python code to consumable output/ data structures. The downside is that it's less quick than doing a dashboard using any of the BI (business intelligence) tools.

BI SOFTWARE

• last time I checked Tableau allows you to use the software for free if you ultimately publish publicly your dashboard

• GCP's dashboarding is also pretty straight forward (i.e. can easily drag and drop)

[bwliv]

Making these hyperparameters shouldn't take more than 20 minutes. I bumped them up to the top of the trainqlearner function - all that's left is to edit sim.py so return_stats takes these same arguments, instead of using a q table and quantiles generated outside of return_stats

[bwliv]

52 will be needed for this too