Hypergraph Interchange Format invitation

[colltoaction]

Hi! I'm Martín and I'm a developer working in the HIF project, an effort to share higher-order datasets developed by members of HyperNetX, XGI and other established higher-order network libraries.

I was just watching Bastien's talk, and found out about MANTRA. I see that the JSON format is important in this library but maybe you can still consider using the HIF standard JSON Schema. We are looking to expand the community (pszufe/HIF-standard#25) and it would be great to have you!

Let me know and I can give you more information. PS: we'll be at NetSci2025.

[Pseudomanifold]

Thanks, that's a great idea, IMO, and I'll discuss it with our team. If I understand the schema correctly, we would have to represent all incidences here, though, so the file size would increase. This could become a problem for hosting the data, but we'll cross that bridge when we arrive on it.

As side-note (I can also open an issue for this), it would be helpful to see an example of a simple hypergraph and its corresponding HIF representation. The repo currently does not contain any examples.

[colltoaction]

Sounds great! I think these tutorials can help in case you haven't seen them:

• https://github.com/pszufe/HIF-standard/blob/main/tutorials/hnx.ipynb
• https://github.com/pszufe/HIF-standard/blob/main/tutorials/xgi.ipynb

I also notice you use the TopoX suite, and we started discussions with them as well. We have a minimal example of interoperability here:

• https://github.com/pszufe/HIF-standard/blob/main/tutorials/tnx.ipynb

Expect more improvements in the docs soon as we write our paper, but here is a snippet about the encoding:

Higher-order interactions in HIF are encoded through incidences, which are structured as lists of records representing edge-node pairs along with any edge-dependent node properties. Similarly, nodes and edges are also organized as lists of records. To accommodate the wide variety of attributes used in different contexts, HIF includes attribute support across nodes, edges, and incidences. However, it is essential to distinguish between structural properties, such as directionality in incidences and weight in nodes and edges, and other attributes. These structural properties are maintained as top-level elements, separate from general attributes.

It would be great to have your feedback and you are welcome to open issues in our repository or send it my way.

[Pseudomanifold]

Thanks, yes, already taking a look at this! We'll be in touch! 😄