{"id":"arxiv/2609.10729","url":"https://arxiv.org/abs/2609.10729","tags":["quant-ph","cs.LG","stat.ML"],"score":0,"title":"A Quantum-Inspired Dequantization Method for Diagonally Weighted Matrix Functions: Application to Learning with Optimized Random Features","domain":"quant-ph","source":"arxiv","authors":["Natsuto Isogai","Mio Murao","Hayata Yamasaki"],"reviews":0,"subject":"quant-ph","summary":"Quantum-inspired classical algorithms have dequantized several quantum machine learning routines by replacing quantum linear-algebra subroutines with classical counterparts. However, the sampler based on quantum singular value transformation (QSVT) for learning with optimized random features is not covered by existing dequantization frameworks, because the matrix to be inverted is not itself available through sampling access. In this work, we develop a classical algorithm to address this type of quantum-advantage candidate. Our method samples heavy indices, reduces the transformation to a…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-09-09","reactions":{}}