Symmetric Jacobi Polynomials on a Triangle and Their Spectral Algebra
arXiv:2607.22751
2026
Architecture
1 ideas extracted · analyzed Aug 30, 2026
What the math gives to ML
The paper constructs a complete orthogonal basis for exchange-symmetric functions of two simplex coordinates using elementary symmetric variables and a nontrivial interaction-weighted geometry. The transferable asset is a principled invariant coordinate system in which polynomial channels are decorrelated under a tunable measure containing the diagonal factor |x-y|^{2\kappa+1}. This suggests an inexpensive pair or edge embedding for graph networks, set models, and attention modules that must be exactly invariant under swapping the two members of a pair. The basis can be estimated once from data and then frozen, giving controlled feature scales and less redundancy than raw symmetric monomials.
Ideas from this paper
Unverified
2026
For every unordered pair of scalar features, construct invariant coordinates from the elementary symmetric quantities s=x+y and q=xy, then feed a truncated orthogonalized polynomial basis in (s,q) to the neural network. Estimate the basis by weighted Gram-Schmidt or Cholesky whitening under the paper's triangle weight, so polynomial channels have low redundancy and controlled scale instead of requiring an unconstrained MLP to learn both symmetry and decorrelation.
Useful5/10
Difficulty3/10
Novelty7/10