A new release 3.2 of the incredible library NetworkX (the Network Analysis tool in Python)
Many improvements including Kemeny's constant on undirected and directed graphs.
🔗 Release notes - https://networkx.org/documentation/stable/release/release_3.2.html
🔵 Notebook meant to be a starting point
🔵 Low code, highly customizable
🔵 Can be converted to an HTML file shareable, emailable
🔵 Interactively explore the elements you want
🔵 Expecting, hoping for suggestions, improvement (code and non-code)
Full video/explanation here (with link to the code and data):
O querido @vepo andou postando um texto massa sobre #grafos: https://blog.vepo.dev/posts/duas-formas-de-pensar-grafos
Isso me lembrou que eu gostaria de continuar lendo o #ThinkComplexity do Downey que é aberto e começa com #Python #networkx:
We're building calculation trees that represent the relationships between different fields in the FERC Form 1 (e.g. how hundreds of granular reported values aggregate up to a utility's balance sheet or income streams) based on XBRL metadata (which we've cleaned up significantly) but the default output leaves something to be desired.
The sx layer in the GeoPKG uses the centre-line track-model overlay on OpenStreetMap #OSM rail data based on minimum point to track distance
The ox uses only OSM data for track and meta-data tags
Noticias sobre Python y Datos de la semana, episodio 73 🐍⚙️
En resumen: Versiones nuevas de JupySQL, Modin, Prefect y más, machine learning "doble", detección de comunidades en grafos, ojo a la chi2 de scikit-learn, y noticias del viejo Spyder
Apoya el noticiero suscribiéndote por correo 📬
1. Don't touch your code for 6 months.
2. Come back, do stuff, push... all tests fail -_-
Over these past 6 months:
- #shapely made their geometries immutable (that was a pain)
- #Networkx decided to move to #scipy sparse arrays, even though they are nowhere near finished and the move was almost guaranteed to break code
- #graph_tool moved a function to another module
- #matplotlib 3.7 made ArtlstList immutable
- #nest broke their ubuntu package
How's *your* day going?
Summer time is open source time. Warm-up: Make sure that Python graph library #NetworkX can calculate minimum weight matchings of bipartite graphs without assuming that the underlying graphs are complete. 
This is useful whenever "something" has to be "assigned" to "someone" where each "someone" has preferences about what they're being assigned.  This turns out to be a common subproblem in combinatorics and constraint satisfaction.