Quantum Multiple Valued Kernel Circuits

Quantum kernels map data to higher dimensions for classification and have been shown to have an advantage over classical methods. In our work, we generalize recent results in binary quantum kernels to multivalued logic by using higher dimensional entanglement to create a qudit memory and show that the use of qudits offers advantages in terms of quantum memory representation as well as enhanced resolution in the outcome of the kernel calculation. Our method is not only capable of finding the kernel inner product of higher dimensional data but can also efficiently and concurrently compute multiple instances of quantum kernel computations in linear time. We discuss how this method increases efficiency and resolution for various distance-based classifiers that require large datasets when accomplished with higher-dimensioned quantum data encodings. We provide experimental results of our qudit kernel calculations with different data encoding methods through the use of a higher-dimensioned quantum computation simulator.

Automated summary

In this study, we generalize recent findings in binary quantum kernels to multivalued logic by utilizing higher dimensional entanglement. Our method offers enhanced resolution and efficiency in kernel calculation, and is capable of concurrently computing multiple instances of quantum kernels. It is suitable for handling large datasets.