SupplementaryMaterial
–Neural Information Processing Systems
We adapt the convergence proof from [46], it can be summarized in two parts. This definition is the same as Def. 1, but for an arbitrary operator. Our final goal is to show that Gα(xi,xj) Dα,Ht(xi,xj), and then conclude with Thm. 1 which gives the equivalence Dα,Ht(xi,xj) d2αM(xi,xj). SincewewillletK, we ignore the second term inGα(xi,xj), as it converges to zero. Bythestronglawoflargenumbers,theconvergence is with probability one.
Neural Information Processing Systems
Feb-11-2026, 17:06:00 GMT
- Technology: