Beyond the Gegenbauer Paradigm: q-Orthogonal Kernels for Machine Learning

The paper proposes an SVM kernel built from discrete q-Hermite I polynomials, using the deformation parameter q to generalize classical Hermite constructions. According to the supplied abstract, the authors establish validity under Mercer’s theorem and argue that boundedness mitigates annihilation and explosion without explicit scaling. Experiments reportedly cover 20 benchmark datasets and show competitive performance against classical and orthogonal-polynomial kernels, with code released on GitHub. However, arXiv identifier 2608.03482 and the stated August 4, 2026 publication date are future-dated, so the paper metadata, proofs, and experimental results cannot currently be independently verified.