QiVC-Net: Quantum-Inspired Variational Convolutional Network, with Application to Biosignal Classification

TL;DR

QiVC-Net introduces a quantum-inspired variational convolution framework that integrates structured uncertainty modeling into CNNs for biosignal classification, specifically phonocardiograms. The key innovation, QiRE sampling, performs norm-preserving unitary rotations of kernel weights within a low-dimensional subspace to generate expressive posterior representations without extra learnable parameters. The architecture combines QiVConv with a reversal fusion residual block to capture bidirectional temporal dynamics, achieving state-of-the-art accuracy (≈97.8–97.9%) on CinC2016 and CirCor 2022 PCG benchmarks while delivering well-calibrated, uncertainty-aware predictions under noise and class imbalance. These results suggest that geometry-preserving probabilistic convolutions can enhance robustness and interpretability in real-world biomedical time-series analysis, with practical implications for reliable clinical decision support.

Abstract

This work introduces the quantum-inspired variational convolution (QiVC) framework, a novel learning paradigm that integrates principles of probabilistic inference, variational optimization, and quantum-inspired transformations within convolutional architectures. The central innovation of QiVC lies in its quantum-inspired rotated ensemble (QiRE) mechanism. QiRE performs differentiable low-dimensional subspace rotations of convolutional weights, analogously to quantum state evolution. This approach enables structured uncertainty modeling while preserving the intrinsic geometry of the parameter space, resulting in more expressive, stable, and uncertainty-aware representations. To demonstrate its practical potential, the concept is instantiated in a QiVC-based convolutional network (QiVC-Net) and evaluated in the context of biosignal classification, focusing on phonocardiogram (PCG) recordings, a challenging domain characterized by high noise, inter-subject variability, and often imbalanced data. The proposed QiVC-Net integrates an architecture in which the QiVC layer does not introduce additional parameters, instead performing an ensemble rotation of the convolutional weights through a structured mechanism ensuring robustness without added highly computational burden. Experiments on two benchmark datasets, PhysioNet CinC 2016 and PhysioNet CirCor DigiScope 2022, show that QiVC-Net achieves state-of-the-art performance, reaching accuracies of 97.84% and 97.89%, respectively. These findings highlight the versatility of the QiVC framework and its promise for advancing uncertainty-aware modeling in real-world biomedical signal analysis. The implementation of the QiVConv layer is openly available in GitHub.

Figures (8)

• Figure 1: Representative PCG signal segments from the CinC dataset, depicting normal and with murmur heart sounds across different recording conditions.
• Figure 2: Representative PCG signal segments from the CirCor dataset, depicting normal and with murmur heart sounds across different recording conditions.
• Figure 3: Reversal fusion residual block architecture.
• Figure 4: Class distribution in the PCG datasets: (a) CinC and (b) CirCor.
• Figure 5: Model robustness on the CinC dataset under varying SNR ratios; (a) AUC vs. SNR, and (b) Accuracy vs. SNR.