Youle Wang

With the continuous development of quantum computing, variational quantum autoencoders (VQA) have become an important model in quantum machine learning, widely applied for the compression and representation of quantum data. However, noise remains a significant challenge in practical quantum computing, especially when quantum circuits are run on simulators and real hardware, where noise can substantially affect the quality of quantum states. To address this issue,inspired by variational quantum autoencoders (VQA) and quantum noise suppression techniques, this paper proposes a novel quantum denoising filter architecture. In this architecture, we introduce auxiliary qubits on top of the variational quantum autoencoder to extend the encoding and decoding capabilities. Both the encoder and decoder are implemented using variational quantum circuits (VQC), which are trained to learn efficient representations of quantum data. To effectively mitigate the impact of noise on quantum information, we incorporate Zero Noise Extrapolation (ZNE) for denoising. This approach combines quantum autoencoding with noise mitigation strategies, providing a new solution for quantum information processing. Experimental results demonstrate that the quantum denoising filter is capable of recovering high-fidelity quantum states in various noisy environments, outperforming traditional noise filtering methods.The core innovation of this work lies in the proposal of a unified denoising mechanism that can simultaneously address both coherent and incoherent noise,enabling higher-quality quantum state recovery and improved model robustness.