Variational quantum deep neural network for image classification
Abstract
<p indent="0mm">Variational quantum algorithms are among the most promising methods to achieve quantum advantage in the NISQ era. The high-dimensional Hilbert space of these algorithms helps capture more complex patterns and features in image data. However, the complex parameter space of variational quantum circuits often leads to issues such as vanishing gradients and local minima, creating a trade-off between trainability and learning capability when dealing with more complex image problems. This paper proposes a variational quantum deep neural network model for image classification. The model is based on a parameterized quantum circuit that is easy to scale, which avoids the vanishing gradient problem and uses efficiently evaluated quantum natural gradients to circumvent local minima. Numerical simulations demonstrate that this model has strong trainability and learning capabilities, enabling it to handle more complex image data.</p>
