Quantum Fidelity Kernel
The quantum kernel is the squared overlap of two encoded states.
Symbols
| $k(x,x')$ | Kernel value |
|---|---|
| $U_\Phi(x)$ | Feature-map circuit |
| $|\Phi(x)\rangle$ | Encoded state |
| $|0^n\rangle$ | All-zero state of n qubits |
$x,x'$ are two samples, $U_\Phi$ is the feature-map circuit, and $|\Phi(x)\rangle=U_\Phi(x)|0^n\rangle$. The second formula is the probability of getting the all-zero string in the compute–uncompute circuit, which is how it is estimated on hardware. $0\le k\le1$ and $k(x,x)=1$, and the kernel is positive semidefinite. It applies to Fidelity Kernels; estimating them requires shots that grow when the values concentrate.
Example
$R_Y(x)$ encoding on one qubit: $k=\cos^2\tfrac{x-x'}{2}$. For $x=0.4$, $x'=1.9$: $\cos^2(0.75)\approx0.5354$. Checking the second formula: $R_Y(-0.4)R_Y(1.9)|0\rangle=R_Y(1.5)|0\rangle$, and the probability of 0 is $\cos^2(0.75)$ (verified).
Test your understanding
The quiz needs JavaScript.
Related topics
References
- Vojtěch Havlíček et al. (2019). Supervised learning with quantum-enhanced feature spaces. Nature 567:209–212 doi:10.1038/s41586-019-0980-2
- Maria Schuld; Nathan Killoran (2019). Quantum Machine Learning in Feature Hilbert Spaces. Phys. Rev. Lett. 122:040504 doi:10.1103/PhysRevLett.122.040504
- Xanadu (2025). Kernel-based training of quantum models with scikit-learn. PennyLane Demos pennylane.ai
