HQCC: A Hybrid Quantum-Classical Classifier with Adaptive Structure

TL;DR

This work tackles the performance bottleneck of fixed-structure PQCs in quantum machine learning by introducing HQCC, a hybrid quantum-classical classifier that adaptively designs PQC architectures via an LSTM-based controller. It combines a local quantum filter (VQCL) with angle-encoded inputs and a sliding-window scheme to enable scalable feature extraction on NISQ devices, while architectural plasticity tunes entanglement depth to balance expressivity and noise resilience. Through experiments on MNIST and Fashion MNIST using TensorCircuit, HQCC demonstrates high classification accuracy and robustness, with the LSTM-enhanced variant (HQCC^b) providing additional gains over non-LSTM configurations. The results suggest that adaptive, memory-enabled circuit design and localized quantum filtering can make quantum methods practical for real-world, large-scale datasets in the NISQ era, albeit with remaining challenges in fine-grained texture discrimination.

Abstract

Parameterized Quantum Circuits (PQCs) with fixed structures severely degrade the performance of Quantum Machine Learning (QML). To address this, a Hybrid Quantum-Classical Classifier (HQCC) is proposed. It opens a practical way to advance QML in the Noisy Intermediate-Scale Quantum (NISQ) era by adaptively optimizing the PQC through a Long Short-Term Memory (LSTM) driven dynamic circuit generator, utilizing a local quantum filter for scalable feature extraction, and exploiting architectural plasticity to balance the entanglement depth and noise robustness. We realize the HQCC on the TensorCircuit platform and run simulations on the MNIST and Fashion MNIST datasets, achieving up to 97.12\% accuracy on MNIST and outperforming several alternative methods.

Figures (3)

• Figure 1: Framework of the QUDM
• Figure 2: Schematic diagram of controller.
• Figure 3: Schematic diagram of controller.