{"id":"arxiv/2610.17722","url":"https://arxiv.org/abs/2610.17722","tags":["quant-ph","cs.LG"],"score":0,"title":"Representation Geometry and Inductive Biases in Quantum Classifiers across Classical Visual Benchmarks","domain":"quant-ph","source":"arxiv","authors":["Yu Yamaoka","Weng Ian Chan","Kosuke Mitarai"],"reviews":0,"subject":"quant-ph","summary":"Quantum Machine Learning (QML) has a theoretical advantage over classical algorithms for specific synthetic data that is computationally hard. However, it remains unclear whether such theoretical advantages play a meaningful role for realistic classical data, such as natural images. Quantum structural properties induced by unitary operations or entanglement introduce inductive biases that differ from those of classical models, potentially leading to distinct learning behavior on real-world data. In a fully quantum learning pipeline, classical real-world data are first encoded into quantum…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-10-13","reactions":{}}