Quantum supervised machine learning with Kerr-nonlinear oscillators

We propose a strategy for obtaining the expressivity in quantum machine learning without the data reuploading, which repeatedly encode into the quantum state. Using Kerr-nonlinear bosonic cavities, we can make full use of a large Hilbert space even if we have a single bosonic mode and it brings highly expressive quantum machine learning. Our numerical simulations show that the expressibility of our method with only one mode is much higher than that of the conventional method with six qubits. Our results pave the way towards a resource efficient algorithm of quantum supervised machine learning.