Memristor-Accelerated Neural Network Architectures for Energy-Efficient Quantum State Tomography

Accurate quantum state tomography (QST) is vital for calibrating quantum processors but faces exponential scaling challenges. We benchmark seven neural architectures—FCN, CNN, CGAN, Transformer, RNN, RBM, and SVAE—for QST reconstruction using expectation- and probability-based measurements. CNN and CGAN achieve high fidelity (F > 0.99), while SVAE enables efficient event-driven learning. To enhance scalability, memristor-based computation-in-memory (CiM) acceleration is proposed for CNN and SVAE, leveraging analog matrix–vector multiplication in HfO2crossbars. The fabricated arrays show stable bipolar switching and STDP behavior, advancing energy-efficient, real-time quantum diagnostics through algorithm–hardware co-design.