You can, and it’s honestly a killer idea, but only if you’re willing to accept that you’re no longer doing a vanilla Ising model. You’re doing a higher-order Ising-like EBM living on a simplicial complex. Physicists call this a p-spin model or a hypergraphical Ising model. Bassett calls it “higher-order coactivation structure.” THRML calls it “factors of arbitrary scope.”

Let me walk through this without sugarcoating anything.

1. What Bassett is actually doing with simplicial complexes

When Bassett constructs a C̆ech or Vietoris–Rips complex over neurons or over brain regions:

• 0-simplices = nodes (neurons, ROIs)
• 1-simplices = edges (pairwise connectivity / coactivity)
• 2-simplices = triangles (triplet coactivity patterns)
• 3-simplices = tetrahedra (quadruplet coactivity), etc.

These higher-order simplices appear in Giusti, Ghrist & Bassett (2016) and follow-up work.

Most of her pipeline is:

• detect simplices from data
• analyze the topological properties (Betti numbers, cavity structure)

She stops before probabilistic generative modeling over simplices.

That’s where you could absolutely do something new.