Quantum Neural Networks: A Comparative Analysis and Noise Robustness Evaluation

Tasnim Ahmed, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique · 2025

Read Quantum Neural Networks: A Comparative Analysis and Noise Robustness Evaluation by Tasnim Ahmed, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-14412.txt.

Introduction

Quantum neural networks promise to blend quantum computing with machine learning, but today's quantum hardware is noisy, and noise can sink a model's accuracy. Ahmed and co-authors compare three hybrid designs and then stress-test the winners.

The comparison covered a quantum convolutional network, a quanvolutional network, and quantum transfer learning, each tried with different circuit layouts, depths, and placements. The best versions then faced five kinds of gate noise, from bit flips to damping. Resilience varied by model and by noise type, but the quanvolutional design proved toughest overall, holding its performance where the others sagged. The authors draw a practical lesson: on noisy machines, the right model depends on the noise it will live with.

Catalog record

CREATORTasnim Ahmed, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-14412

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Quantum Neural Networks: A Comparative Analysis and Noise Robustness Evaluation Tasnim Ahmed, Muhammad Kashif, Alberto Marchisio, Muhammad Shafique Abstract In current noisy intermediate-scale quantum (NISQ) devices, hybrid quantum neural networks (HQNNs) offer a promising solution, combining the strengths of classical machine learning with quantum computing capabilities. However, the performance of these networks can be significantly affected by the quantum noise inherent in NISQ devices. In this paper, we conduct an extensive comparative analysis of various HQNN algorithms, namely Quantum Convolution Neural Network (QCNN), Quanvolutional Neural Network (QuanNN), and Quantum Transfer Learning (QTL), for image classification tasks. We evaluate the performance of each algorithm across quantum circuits with different entangling structures, variations in layer count, and optimal placement in the architecture. Subsequently, we select the highest-performing architectures and assess their robustness against noise influence by introducing quantum gate noise through Phase Flip, Bit Flip, Phase Damping, Amplitude Damping, and the Depolarizing Channel. Our results reveal that the top-performing models exhibit varying resilience to different noise gates. However, in most scenarios, the QuanNN demonstrates greater robustness across various quantum noise channels, consistently outperforming other models. This highlights the importance of tailoring model selection to specific noise environments in NISQ devices. 1 Introduction Quantum Computing (QC) is a new computational paradigm that can potentially solve computationally intensive problems substantially faster than its classical counterpart. Over the last few years, there has been significant progress towards the development of quantum computers and the Noisy-Intermediate Scale Quantum (NISQ) devices [ 1 ] have already been developed. These devices have a limited number of qubits and demonstrate limited resilience to noise. Despite the limitations, the NISQ devices are already being used to demonstrate the so-called quantum advantage for various applications [ 2 , 3 , 4 ] . The development of NISQ devices has opened avenues to explore a wide range of applications for the post-quantum era. One such application is Quantum Machine Learning (QML) [ 5 , 6 , 7 , 8 ] . QML combines the field of Classical Machine Learning (ML) and QC, and aims to benefit from the development of advanced classical ML models, architectures, training methodologies, and infrastructure to build ML algorithms, such as Deep Neural Networks (DNN) in conjunction with the distinctive computational capabilities of quantum computers [ 9 , 10 , 11 ] . Inspired by the tremendous success of classical DNNs, the Quantum Neural Networks (QNNs) stand out as one of the most studied QML algorithms [ 12 ] . However, due to the limitations of the NISQ devices, the development and practical realization of standalone QML algorithms including QNNs is challenging. This has lead to the development of hybrid quantum-classical neural networks (HQNNs), which are designed to be NISQ-compatible and are notably robust against errors typical in NISQ devices. The fundamental part of an HQNN is the Variational Quantum Circuit (VQC), which utilizes parameterized quantum gates. The parameters of these gates are trainable which are optimized during the training using classical methods [ 13 ] . Figure 1: Motivational Case study. (a)Two different variant of HQNN (QCNN and QuanNN) yields different performance with the same underlying architecture of quantum layers (b) Different effects of Bit Flip noise on the performance of QCNN and QuanNN highlight unique noise sensitivities of different HQNNs. HQNNs have garnered significant interest due to their expressiveness and reduced trainable parameters, made possible by the VQC [ 14 ] . These networks have been proposed for diverse applications such as generative modeling [ 15 , 16 ] , classification [ 17 ] , and regression analysis [ 18 ] , thus prompting the development of various HQNN algorithms. Examples include Quantum Recurrent Neural Networks, Quantum Reinforcement Learning Models, and Quantum Graph Neural Networks, each contributing to the broad QML landscape. This paper examines three frequently used HQNN models used in image classification tasks: Quanvolutional Neural Networks (QuanNN) [ 19 ] , Quantum Convolutional Neural Networks (QCNN) [ 20 ] , and Quantum Transfer Learning (QTL) [ 21 ] , focusing on their unique architecture and performance. Each of these models uses quantum circuits uniquely to leverage quantum mechanics for machine learning tasks. Both QuanNN and QCNN draw inspiration from classical Convolutional Neural Networks. The QuanNN [ 19 ] uses a single quantum circuit as a filter which slides over the entire image and extract the features which can then be trained using either quantum circuits or classical layers. The QCNN, on the other hand, downsizes the input image to match the first circuit’s qubit count and performs pooling by reducing the number of qubits in successive circuits until reaching the desired output size [ 20 ] . A key difference between QuanNN and QCNN is that the QuanNN does not have a pooling operation. The QTL model is inspired by classical transfer learning. It involves transferring knowledge from a pre-trained classical network to a quantum setting, where a quantum circuit is integrated for quantum post-processing [ 21 ] . Numerous studies have been conducted to explore the circuit architecture and the trainability challenges for the above-discussed algorithms [ 22 , 23 , 24 , 25 ] . However, a comprehensive study that compares the performance of these algorithms for a specific application is still lacking. Additionally, most of the state-of-the-art studies primarily focus on their performance in ideal conditions and do not take into consideration the effect of different quantum noise, which is an inherent characteristic of NISQ devices. 1.1 Related Work The work in [ 22 ] investigates the performance of various QCNN models, which differed in their structures of parameterized quantum circuits, quantum data encoding methods, and classical data preprocessing approaches when applied to MNIST datasets. That study reported that QCNNs achieved high classification accuracy despite a limited number of free parameters. However, their analysis was limited to noise-free model performance. Whereas, our work evaluates the best performing QCNN model under various quantum noise channels, investigating whether or not the model exhibits robustness to noise over time . An extensive comparative analysis of different QNN algorithms, including QuanNN, QCNN, and Quantum Residual Network, is presented in [ 23 ] . The paper examined the effects of varying circuit depths, the number of qubits, and different entanglement settings on image classification tasks using the MNIST dataset. Although the paper provided a thorough evaluation of different models under varying circuit architectures, it did not analyze the performance implications of noise on these models. In this paper we expand upon this analysis by examining the robustness of the best-performing model under different noise channels. The impact of 5 different noise channels on the trainability of HQNNs is investigated in [ 24 ] . Their study employed a two-qubit VQC within the HQNN architecture, focusing on its performance in binary classification tasks. In contrast, our paper conducts a similar analysis with the same noise channels but extends it to larger 4-qubit circuits for multiclass classification tasks on MNIST, highlighting the robustness of larger circuits against noise . These studies offered substantial insights into the performance of various models under ideal conditions but provided minimal understanding of the interplay between noise and circuit architecture. Therefore, we believe that a thorough exploration of different noise channels associated with

Read the complete text