pygraphistry

This! We do it all the time in fraud, genomics, social media, security, etc data tables, it is one of our favorite techniques. We do one more step for UMAP and other dimensionality reducers in practice: connect the nearest neighbors to turn any data table into an explorable similarity graph. Users, devices, proteins, accounts, events, etc link together and cluster when similar, and this reveals fascinating structural relationships when there is a reason they vary. Takes just a few lines once you have a feature table, and there's a recent GH issue to automate that step too: https://github.com/graphistry/pygraphistry/blob/master/demos...

The duality is super useful in practice! In the table/dataframe -> hypergraph dataframe transform @ https://github.com/graphistry/pygraphistry , we do `hypergraph(multicolumn_table, direct=True | False)['graph'].plot()` , which renders hypergraphs as a regular graph. That lets you pick. Saves a lot of wrangling! Consider exploring some logs of customer activity or security events. A hyperedge becomes either: - a node of a bipartite graph. Ex: each log event becomes a node connecting the various entity nodes it mentions. Event <> IPs, accounts, countries, ... - .. or a bunch of pairwise entity<>entity edges. Ex: connect each IP<>account<>country directly, and label each edge with the hyperedge event id it came from. In both cases, you can now directly leverage a lot of traditional graph thinking, and in our case, GPU acceleration & visual

https://github.com/graphistry/pygraphistry + https://github.com/rapidsai/cudf if you're coming from the api / ds side, and GNNs side is more of a "let's chat!" thing for figuring out aligned collaborations (slack/email, all good!)

Yep; fast+safe+well-tooled common IO format for tabular data, especially columnar, for going across runtimes/languages/processes/servers with zero copy. It supports a variety of common types, including nested (ex: lists, records, ...), modes like streaming, and schemas that can be inferred+checked, so it covers a lot of classic protobuf messaging use cases. When messages look more like data, even better, as columnar representations are often better for compression & compute. So it's pretty common now for data frameworks, and increasingly, DB connectors. Ex: Our client https://github.com/graphistry/pygraphistry takes a bunch of formats, but we convert to Arrow before sending to our server, we zero-copy it between server-side GPU processes, and stream it to WebGL buffers in the client, all without touching it (beyond our own data cleaning).