Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

QI: Fachverband Quanteninformation

QI 10: Quantum Machine Learning II

QI 10.6: Vortrag

Dienstag, 11. März 2025, 12:15–12:30, HS VIII

Quantum Support Vector Machines Kernel Generation with Classical Post-Processing — •Anant Agnihotri, Thomas Wellens, and Michael Krebsbach — Fraunhofer IAF, Tullastrasse 72

We investigate the optimization of kernel generation for quantum support vector algorithms for data classification. Classical post-processing techniques are employed to improve the efficiency of classification. First, high-dimensional data is preprocessed using Principal Component Analysis (PCA) to reduce dimensionality while retaining significant features. A training kernel is then generated using the ZZ feature map. In the post-processing step, the overlap with all states (not only the all-zero state, as it is the case for the standard quantum kernel) is utilized, where the kernel entry is computed as a weighted sum of these overlaps. This allows us to determine the kernel entries with reduced number of shots. The method is run on MNIST dataset to distinguish between handwritten digits *0* and *1*. We compare the kernel score, i.e., the fraction of unseen datapoints correctly identified by the standard quantum kernel, on the one hand, and the kernel with our post-processing method, on the other hand. Our findings indicate that the post-processed version outperforms the standard version especially for higher numbers of qubits.

Keywords: Quantum Machine Learning; Support Vector Machines; Quantum Kernel