Variational Quantum Anomaly Detection: Unsupervised mapping of phase diagrams on a physical quantum computer

Code, simulations and real-device experiments for our preprint arXiv:2106.07912
Entry for the Qiskit Europe Hackathon

We propose variational quantum anomaly detection (VQAD), a novel quantum machine learning framework for exploring phase diagrams of quantum many-body systems. VQAD is trained in a fully unsupervised fashion on a quantum device. The implentation is done with Qiskit. We walk you through our proposal in main.ipynb.

• qae.py: Includes QAEAnsatz, the parameterized circuit ansatz for the QAD, and QAE, the training framework
• experiments/: The experiments performed on the physical device. Each file has a short description of the experiment in the beginning. The notebook that produced the results shown in the paper is jakarta_antiferro_execute.ipynb.
• main.ipynb: A complete run through our proposal showcasing all its ingredients. In its form here it is generating the result for the 2D Antiferromagnetig Ising model in real-noise simulations
• simulations/qae-from-vqe.ipynb: The QAE results for the BH model
• data/: Our notebooks are configured in the way that all results are stored in numpy files for reuse
• plotting/: Notebooks to replot the results in the desired fashion for the presentation and paper