Installation

Requirements

This Not That (TNT) is built on top of (and as an extension of) the Panel library. Thus at a minimum you will need to have panel installed to make use of TNT. Panel itself is built atop Bokeh, and Bokeh is the best default choice for plotting data maps within TNT. To make effective use of TNT you will also need numpy and pandas installed. Given that you will be exploring data maps you will likely want to install umap-learn anyway, which will bring in numpy, scikit-learn, numba and pynndescent (which will all be useful with TNT anyway). For color handling matplotlib is needed, and finally for cluster label annotations hdbscan is required.

In summary you will need:

• panel + bokeh + param

• umap-learn + numpy >= 1.22 + scikit-learn + numba + pynndescent

• pandas

• matplotlib

• hdbscan

• cmocean

• colorcet

• glasbey

• cmocean

• vectorizers

Recommended Extras

While the requirements are enough to get you working with TNT, there are some optional extras that can prove to be very beneficial. If you are interested in text annotations of clusters in data maps then both apricot-select for submodular selection, and networkx for graph handling can be very beneficial.

• apricot-select

• networkx

Installing via pip

Currently you can pip install directly from this repository:

pip install git+https://github.com/TutteInstitute/thisnotthat

We hope to have a beta version of TNT on PyPI soon, and versions available on conda-forge sometime following that.