Abstract:
Image classification remains a fundamental problem in artificial intelligence, with applications
spanning medical diagnostics, autonomous systems, and security surveillance.
Traditional deep learning models, such as Convolutional Neural Networks (CNNs), have
achieved state-of-the-art performance in image classification tasks. However, these models
suffer from high computational costs and limitations in optimization efficiency. Recently,
quantum computing has emerged as a promising alternative, particularly in hybrid
quantum-classical models that leverage quantum computational advantages for improved
learning. This study introduces a novel Optimization-Based Image Classification
Framework using a Time-Nonlocal Optimization Approach. Our method integrates
time-dependent parameterization to smooth gradient variations, thereby mitigating
the barren plateau problem commonly observed in quantum machine learning. By leveraging
time-nonlocal optimization, we enhance the training stability and convergence of
quantum variational models, making them more viable for high-dimensional classification
tasks. We evaluate our proposed framework on benchmark datasets, including Iris and
MNIST, to demonstrate its effectiveness. Experimental results show that our approach
significantly improves classification accuracy while reducing training inefficiencies compared
to standard quantum optimization techniques. The proposed method also offers
improved scalability, making it a practical solution for hybrid quantum-classical image
classification. This research contributes to the advancement of quantum-enhanced image
classification by addressing key optimization challenges and demonstrating the feasibility
of time-nonlocal optimization in quantum-classical hybrid learning. The findings of this
work provide a foundation for future developments in quantum-assisted deep learning and
optimization-based image processing.
