Abstract
Quantum feature selection offers a powerful route to extracting informative features from high-dimensional data by exploiting quantum parallelism and superposition. However, most existing methods focus on single-view data and fail to address the challenges of multi-view scenarios, where heterogeneous views must be integrated while maintaining the fidelity of each view. In such scenarios, achieving both inter-view complementarity and intra-view consistency is difficult due to complex dependencies and latent structures within each view. To address this problem, we propose Quantum Multi-View Feature Selection (QMVFS), a unified framework that combines configurable quantum kernel circuits with adaptive multi-kernel fusion. For each view, configurable quantum kernel circuits with distinct architectures and parameterizations embed data into a Hilbert space, producing diverse kernel representations. An integrated optimization objective jointly learns view weights, fuses multiple kernels, and enforces sparsity, enabling QMVFS to capture inter-view synergy while preserving intra-view structures. Experiments on six public multi-view datasets show that QMVFS outperforms state-of-the-art baselines by an average of 4.44% in classification accuracy, demonstrating its effectiveness and adaptability for quantum-enhanced multi-view learning. Source code is available at: https://github.com/xhchangsha/Quantum-Multi-View-Feature-Selection.
Funding Information
  • National Key Research and Development Project (2022YFB2703100)
  • Postdoctoral Fellowship Program of China Postdoctoral Science Foundation (GZC20251062)
  • China Postdoctoral Science Foundation (2025M781521)

This publication has 53 references indexed in Scilit: