Local and Multi-Scale Strategies to Mitigate Exponential Concentration in Quantum Kernels

Claudia Zendejas-Morales, Debashis Saikia, Utkarsh Singh · 2026

Read Local and Multi-Scale Strategies to Mitigate Exponential Concentration in Quantum Kernels by Claudia Zendejas-Morales, Debashis Saikia, Utkarsh Singh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-16097.txt.

Description

Claudia Zendejas-Morales leads a team of three researchers fighting a fade-out problem in quantum kernels. Fidelity-based quantum kernels link quantum feature maps to classical kernel methods, but as systems grow or circuits grow more expressive, the similarity matrix drifts toward the identity and informative structure washes out. This exponential concentration threatens the practical use of such kernels.

The team tests two fixes built in Qiskit: local kernels that aggregate similarities patch by patch, and multi-scale kernels that mix local and global similarity across patch sizes. Benchmarks on several tabular data sets, under matched preprocessing and SVM protocols, show both fixes ease concentration and enrich kernel spectra versus the global baseline. Effects on classification accuracy vary with data set and feature dimension, so mitigation helps structure even when accuracy gains depend on context.

Catalog record

CREATORClaudia Zendejas-Morales, Debashis Saikia, Utkarsh Singh
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-16097

Browse the full library · About the Tower of Babel