Realizing quantum convolutional neural networks on a superconducting quantum processor to recognize quantum phases.

Area of Science:

• Quantum Information Science
• Condensed Matter Physics
• Machine Learning

Background:

• Characterizing quantum states is vital for quantum computing but becomes computationally intensive with system size.
• Conventional measurement techniques struggle with scalability and cost for large quantum systems.
• Quantum neural networks offer a promising alternative for efficient quantum state recognition.

Purpose of the Study:

• To implement a quantum convolutional neural network (QCNN) for identifying symmetry-protected topological (SPT) phases.
• To benchmark the QCNN's performance against conventional measurement methods on a real quantum processor.
• To explore the potential of QCNNs in characterizing complex quantum states and phases.

Main Methods:

• A 7-qubit superconducting quantum processor was utilized to realize the QCNN.
• Approximate ground states of cluster-Ising Hamiltonians were prepared using a hardware-efficient circuit.
• The QCNN's ability to recognize SPT phases was evaluated by measuring a non-zero string order parameter.

Main Results:

• The QCNN successfully identified the topological phase of the prepared quantum states.
• The QCNN demonstrated higher fidelity in phase recognition compared to direct measurements of the string order parameter.
• The QCNN's performance was robust despite the presence of finite-fidelity gates in its own operations.

Conclusions:

• Quantum convolutional neural networks are effective tools for identifying topological phases in quantum systems.
• QCNNs offer a more efficient and higher-fidelity approach than conventional methods for quantum state characterization.
• This work highlights the potential of hybrid quantum-classical machine learning models in advancing quantum information science.