Q&AIntermediatequantum machine learning

What is a quantum kernel, and is it better than an RBF kernel?

A quantum kernel is the overlap |⟨φ(x)|φ(x′)⟩|² of encoded states; it is not better in general and may suffer exponential concentration.

A quantum feature map $x\mapsto|\phi(x)\rangle=U(x)|0\rangle$ defines the fidelity kernel $k(x,x')=|\langle\phi(x)|\phi(x')\rangle|^2$, which is estimated on a quantum device and then passed to a classical SVM. In the lab it is computed exactly from the state vector, and the ZZ map matches Qiskit's zz_feature_map.

Is it better? There is no general answer. Synthetic examples have been designed in which quantum kernels outperform under complexity-theoretic assumptions, but for typical real data, RBF kernels and classical models are highly competitive (Huang et al. 2021). As the number of qubits grows, kernel values may concentrate around a fixed value, so that a huge number of measurements is needed to distinguish them (Thanasilp et al. 2024). The lab shows the spread of the off-diagonal values as a diagnostic for this, and compares with RBF and linear kernels on the same data.

Related topics

References

  1. Vojtěch Havlíček et al. (2019). Supervised learning with quantum-enhanced feature spaces. Nature 567:209–212 doi:10.1038/s41586-019-0980-2
  2. Maria Schuld; Nathan Killoran (2019). Quantum Machine Learning in Feature Hilbert Spaces. Phys. Rev. Lett. 122:040504 doi:10.1103/PhysRevLett.122.040504
  3. Hsin-Yuan Huang et al. (2021). Power of data in quantum machine learning. Nature Communications 12:2631 doi:10.1038/s41467-021-22539-9
  4. Supanut Thanasilp; Samson Wang; M. Cerezo; Zoë Holmes (2024). Exponential concentration in quantum kernel methods. Nature Communications 15:5200 doi:10.1038/s41467-024-49287-w
  5. Xanadu (2025). Kernel-based training of quantum models with scikit-learn. PennyLane Demos pennylane.ai