<![CDATA[The advancement of quantum computing has encouraged the use of Variational Quantum Neural Networks (VQNNs) as an alternative approach for data classification tasks. However, the implementation of VQNNs in the Noisy Intermediate-Scale Quantum (NISQ) era still faces major challenges due to quantum noise, which can significantly degrade model performance. This study aims to analyze the robustness of VQNNs against quantum noise in a binary classification task using a subset of the MNIST dataset. Image data are processed through normalization and dimensionality reduction using Principal Component Analysis (PCA), and subsequently encoded into quantum circuits using the angle encoding method. The VQNN model is designed with six qubits and a single variational layer, and evaluated under two conditions: an ideal quantum simulation (before noise, statevector) and a noisy quantum simulation (after noise). Quantum noise is simulated using Qiskit Aer, considering two types of noise, namely bit flip and depolarizing noise, across various probability levels. Experimental results show that under ideal conditions, the VQNN achieves a test accuracy of 85.2%. However, increasing noise probabilities lead to a gradual degradation in performance, approaching the random classification threshold at high noise levels. Bit flip noise causes a more severe performance decline compared to depolarizing noise. These findings confirm that quantum noise is a critical factor affecting the stability of VQNNs and highlight the importance of developing effective noise mitigation strategies for VQNN implementations in the NISQ era.]]>
Copyrights © 2026
