add resources for network visualization

[awillats]

as a subset of this, would be good to point to resources for rendering probabilistic graphical models (PGM) :

• daft-pgm

  Daft is a Python package that uses matplotlib to render pixel-perfect probabilistic graphical models for publication in a journal or on the internet

• tikz-bayesnet nicer interface to TikZ for graphical models

  • see also Graphical Models for Causal Inference using LaTeX by Eleanor Murray

• mermaid-js

  • if you're okay relinquishing control of node placement, mermaid can often get you something that looks reasonable quickly with (arguably) nicer syntax than TikZ. For complex diagrams, it becomes more obvious that it's flexible flowcharting software rather than a dedicated graphical model tool.
  • can be embedded in markdown and notebook environments (jupyter / hydrogen)

• dagre-js - directed graph visualization

  • NOTE: dagre and dagre-js are deprecated, but it looks like the same organization is actively maintaining a newer project: graphlib
  • example in observable by Batosz Prusinowski

for rendering machine learning models (esp. artificial neural networks) as animations, see helblazer811/ManimML built on Manim, a Python library for mathematical animations

[awillats]

Some network visualization libraries, resources

• NetworkX

  • largely, a library for analyzing networks, but does have several methods for drawing them also
  • some drawing methods call GraphViz under the hood to handle layout ...
  • nx_altair "nx_altair offers a similar draw API to NetworkX but returns Altair Charts instead."
  • see awillats/clinc/.../network_plotting_functions.py and awillats/clinc/.../svg_draw.py (work in progress) for converting from networkX specification to SVG via Python
  • see awillats/clinc/.../mermaid_str_to_networkx() for parsing a mermaid-js syntax network string to a NetworkX object

• GraphViz

  • uses DOT as language for specifying graphs
  • several options for python interfaces graphviz pygraphviz pydot
  • magjac/d3-graphviz - "Graphviz DOT rendering and animated transitions using D3"

• D3.js - data driven documents in javascript

  • powerful widely used library for creating visualizations via HTML, SVG, and CSS
  • does much more than network viz, but that also means having to dip your toe into a large ecosystem
  • example
  • Altair provides an (indirect) Python interface to d3 (via Vega-Lite)
  • using d3 to render networks in the style of London underground maps [johnwalley/d3-tube-map], stereobooster/d3-tube

• scikit-network - can visualize graphs as simple SVG

  • igraph, python-igraph, [igraph] in R](https://igraph.org/r/)
    • like NetworkX, this is primarily a network analysis library, but has some plotting methods
    • Network Analysis and Visualization with R and igraph detailed blogpost by Katherine Ognyanova

Interactive options

• pyvis

  • built around vis-js

• via Plotly/Dash

  • using cytoscape via dash - dash-cytoscape

  • using d3.js via dash - dash-network

  • awillats/circuit-visualizer-p5

    • my work, very much a work in progress
    • currently built as an application-specific visualizer, not yet general purpose

• networkD3 in R

  • example uses igraph to construct network

• Jaal "Python based interactive network visualizing tool built using Dash and Visdcc"

Comparisons / more resources

• Visualizing Networks in Python by Mohit Mayank

  • discusses networkx, pyviz, and dash

• Network visualization with R by Katherine Ognyanova

  • see also Static and dynamic network visualization with R (blogpost) , and associated repo

Standalone applications / web interfaces

• Gephi (standalone app)

  • contains many interesting network datasets
  • has lots of layout options, including "force atlas"

• Cytoscape (standalone app)

  • highly customizable
  • has a dash interface [docs] which can be called from Python

• Wolfram GraphPlot