Table 2.

Comparison of the cross-validation AUC performance for different data encodings.

Dataset	qDC	qKSVM	sqKSVM	sqKSVM*	qubits
Pediatric Bone Marrow Transplant 2YS	0.62	0.63	0.62	0.61	${\log }_{2}N$
	0.61	0.63	0.61	0.59	$N$
Wisconsin Breast Cancer Malign-vs-benign	0.92	0.92	0.88	0.87	${\log }_{2}N$
	0.90	0.91	0.87	0.85	$N$
Heart failure Mortality	0.62	0.51	0.51	0.50	${\log }_{2}N$
	0.60	0.51	0.51	0.50	$N$

The qDC, qKSVM, and sqKSVM run on Pennylane simulator for $N=8$. For the ${\log }_{2}N$ encoding, $N$ features are encoded into ${\log }_{2}N$ qubits with sequences of Pauli-Y gate rotations ($R_y$) and $\mathit{CNOT}$ s. In another strategy, $N$ features are encoded into $N$ qubits with sequences of the Hadamard gates, Pauli-Z gate ($R_z$) rotations followed by nearest neighbor $\mathit{CNOT}$ s.

*The sqKSVM was also executed on the IBMQ Melbourne machine for reference comparison.