← All problems

Deep Sprecher Network universality conjecture

1 year old · traced to

Let n≥1n\geq 1, L≥1L\geq 1, and let f:[0,1]n→Rf:[0,1]^n\to\mathbb{R} or f:[0,1]n→Rmf:[0,1]^n\to\mathbb{R}^m be continuous. Consider a Sprecher Network with architecture n→[d1,…,dL]→1n\to[d_1,\dots,d_L]\to 1 or →m\to m, with continuous splines ϕ(ℓ),Φ(ℓ)\phi^{(\ell)},\Phi^{(\ell)}, suitable parameters λ(ℓ),η(ℓ)\lambda^{(\ell)},\eta^{(\ell)}, and optionally lateral mixing parameters τ(ℓ),ω(ℓ)\tau^{(\ell)},\omega^{(\ell)}. Deep Sprecher Network universality conjecture. For every ϵ>0\epsilon>0, sufficiently large hidden widths d1,…,dLd_1,\dots,d_L—possibly satisfying dℓ≥2dℓ−1+1d_\ell\geq 2d_{\ell-1}+1, though this condition is not asserted to be necessary—allow the network output f^(x)\hat f(\mathbf{x}) to satisfy

sup⁡x∈[0,1]n∣f(x)−f^(x)∣<ϵ,\sup_{\mathbf{x}\in[0,1]^n}|f(\mathbf{x})-\hat f(\mathbf{x})|<\epsilon,

with the corresponding vector-norm condition in the Rm\mathbb{R}^m case.

The conjecture concerns whether stacking multiple Sprecher blocks preserves universal approximation and whether depth or lateral mixing improves efficiency or learning dynamics. The supplied context says these properties remain unexplored, so the conjecture is open.

References
  1. Christian Hägg, Kathlén Kohn, Giovanni Luca Marchetti and Boris Shapiro, “Sprecher Networks: A Parameter-Efficient Kolmogorov-Arnold Architecture”, arXiv:2512.19367 (2026).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.