Move visualization functions to plotly and get rid of matplotlib dependency

[fealho]

Resolve #348.

A notebook testing the new plots can be found here.

[codecov-commenter]

Codecov Report

Attention: 92 lines in your changes are missing coverage. Please review.

Comparison is base (64a91e8) 87.64% compared to head (ddca46c) 86.85%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #368      +/-   ##
==========================================
- Coverage   87.64%   86.85%   -0.79%     
==========================================
  Files          27       27              
  Lines        1772     1788      +16     
==========================================
  Hits         1553     1553              
- Misses        219      235      +16     

Files	Coverage Δ
copulas/visualization.py	0.00% <0.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[frances-h]

Can you add some screenshots showing the new plots?

[fealho]

Can you add some screenshots showing the new plots?

@frances-h Added to the description.

[amontanez24]

Task linked: CU-860qhuhh5 Copulas - Move visualization functions to plotly and get rid of matplotlib dependency #348

[npatki]

Can you add some screenshots showing the new plots?

@frances-h Added to the description.

@fealho they are not rendering for me. I think that if you upload a local ipynb notebook into Drive, then Google Colab won't render the graphics. Is it possible to create a notebook directly on Google Colab and run the code from a branch or something? Doing in Colab would be nice bc we can also verify the the interactivity that all plotly graphs come with.

[npatki]

I'm leaving all my feedback here based on the notebook.

Plot Titles

Original issue had mentioned we should update any titles to accurately reflect what is being shown. I didn't see a title for any of the generated plots, we should add that.

I'm basing the proposed titles based off the SDMetrics visualizations, which is what we wanted to align to.

• dist_1d: Data for column '<name>'. If there is no name (eg. you're passing in a Series without a name, we could leave that out)
• compare_1d: Real vs. Synthetic Data for column '<name>' (again can leave name out if there is none)
• scatter_2d: Data for columns '<name>' and '<name>'
• compare_2d: Real vs. Synthetic Data for columns '<name>' and '<name>'
• scatter_3d: Data for columns '<name>', '<name>' and '<name>'
• compare_3d: Real vs. Synthetic Data for columns '<name>', '<name>' and '<name>'

X and Y Axes Labels

For the 1D plots (dist_1d and compare_1d) the axes should be labeled. These can be the same as SDMetrics. (X-axis is value and Y-axis is frequency)

[frances-h]

Looks good!

[npatki]

Approved the title and axes changes. Though we should probably update that first notebook as a sanity check.