Emerging generalization advantage of quantum-inspired machine learning in the diagnosis of hepatocellular carcinoma

Research into quantum advantage is increasingly taking on an interdisciplinary character. In particular, quantum machine learning shows promising generalization capabilities, which we have exploited in the classification of hepatocellular carcinoma tissue based on microarray gene expressions. By using previously characterized genetic communities, we minimize the computational complexity associated with the number of qubits, enabling the execution of quantum-inspired algorithms on classical machines. We consider two categories of such algorithms: parameterized quantum circuits (PQC) and tensor networks. The variational optimization of PQCs achieves better accuracy than classical counterparts on the independent test set, reaching an advantage equal to 11% in accuracy, while tensor networks offer equivalent performance with fewer parameters.