eprintid: 72177
rev_number: 10
eprint_status: archive
userid: 12460
dir: disk0/00/07/21/77
datestamp: 2025-07-25 08:24:22
lastmod: 2025-07-25 08:24:22
status_changed: 2025-07-25 08:24:22
type: thesis
metadata_visibility: show
contact_email: muh.khabib@uin-suka.ac.id
creators_name: Supardi Atisina, NIM.: 23206051024
title: DESAIN HYBRID QUANTUM CONVOLUTIONAL NEURAL NETWORKS (QCNN) DALAM PENINGKATAN PERFORMANCE PADA KLASIFIKASI CITRA
ispublished: pub
subjects: 004.
divisions: S2_inf
full_text_status: restricted
keywords: Quantum Convolutional Neural Networks (QCNN), Quantum Machine Learning (QML), Klasifikasi Citra, Komputasi Kuantum, Convolutional Neural Network (CNN)
note: Prof. Dr. Ir. Shofwatul 'Uyun, S.T., M.Kom., IPM., ASEAN Eng
abstract: Convolutional Neural Networks (CNNs), the standard in image classification, face fundamental challenges including a large parameter requirement which is computationally expensive, and limitations in capturing non-local data correlations. As an alternative paradigm, Quantum Convolutional Neural Networks (QCNNs) theoretically offer a solution through higher representational capacity with a more efficient architecture. This research aims to implement and quantitatively evaluate a hybrid QCNN model, and to comprehensively compare its performance against classical CNN architectures.
This research employed an experimental method on a 3-class MNIST dataset reduced to a 4x4 pixel resolution. The hybrid QCNN model was designed using TensorFlow Quantum with a 16 data-qubit and 3 readout-qubit architecture, utilizing Ising-type interactions for feature extraction. Experimental results show that the QCNN achieved a superior accuracy of 70.67%, significantly outperforming the CNN models (~49% accuracy). Furthermore, the QCNN proved to be 41% more parameter-efficient (108 vs. 183), albeit with a significant computational challenge on the classical simulator.
It is concluded that the hybrid QCNN architecture has significant potential to improve classification accuracy and model efficiency compared to simple CNN architectures for this task, thus validating its theoretical advantages. The computational challenges arising from simulation underscore the importance of future quantum hardware development. Therefore, future research is recommended to focus on testing the model on more complex datasets and implementing it on real quantum hardware for real-world performance validation.
date: 2025-07-14
date_type: published
pages: 83
institution: UIN SUNAN KALIJAGA YOGYAKARTA
department: FAKULTAS SAINS DAN TEKNOLOGI
thesis_type: masters
thesis_name: other
citation:   Supardi Atisina, NIM.: 23206051024  (2025) DESAIN HYBRID QUANTUM CONVOLUTIONAL NEURAL NETWORKS (QCNN) DALAM PENINGKATAN PERFORMANCE PADA KLASIFIKASI CITRA.  Masters thesis, UIN SUNAN KALIJAGA YOGYAKARTA.   
document_url: https://digilib.uin-suka.ac.id/id/eprint/72177/1/23206051024_BAB-I_IV-atau-V_DAFTAR-PUSTAKA.pdf
document_url: https://digilib.uin-suka.ac.id/id/eprint/72177/2/23206051024_BAB-II_sampai_SEBELUM-BAB-TERAKHIR.pdf