Deep quantum neural networks on a superconducting processor

This paper experimentally demonstrates training deep quantum neural networks (DQNNs) using the backpropagation algorithm on a six-qubit superconducting processor. Three-layer DQNNs efficiently learned two-qubit quantum channels (96.0% mean fidelity) and the ground state energy of molecular hydrogen (93.3% accuracy). Six-layer DQNNs achieved 94.8% mean fidelity for learning single-qubit channels. The number of coherent qubits needed doesn't scale with DQNN depth, offering valuable guidance for quantum machine learning applications.

Xiaoxuan Pan, Zhide Lu, Weiting Wang, Ziyue Hua, Yifang Xu, Weikang Li, Weizhou Cai, Xuegang Li, Haiyan Wang, Yi-Pu Song, Chang-Ling Zou, Dong-Ling Deng, Luyan Sun