Wijaya, Ridho Nur Rohman (2026) Konstruksi Lapisan Ekstraksi Fitur dan Routing Berbasis Kuantum Pada Model Capsule Network untuk Klasifikasi Citra. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 7002231002-Doctoral.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
7002231002-Doctoral.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy |
Abstract
Klasifikasi citra merupakan bidang penelitian yang penting dalam kemajuan teknologi dan bermanfaat pada kehidupan sehari-hari, seperti sistem deteksi kecelakaan dan diagnosis medis. Capsule Network adalah salah satu model deep learning yang digunakan untuk mengatasi masalah klasifikasi citra dengan keunggulan menangkap informasi spasial melalui algoritma routing. Akan tetapi, model ini memiliki keterbatasan pada teknik ekstraksi fitur yang sederhana dan besarnya beban komputasi pada proses routing. Para peneliti telah menunjukkan bahwa komputasi kuantum dapat meningkatkan kinerja model deep learning dengan memanfaatkan sifatsifat mekanika kuantum. Pada disertasi ini, telah dilakukan konstruksi lapisan ekstraksi fitur dan routing berbasis kuantum pada model Capsule Network. Komponen ekstraksi fitur dikembangkan menggunakan lapisan sirkuit kuantum (Variational Quantum Circuit) dengan dua tahapan, yaitu quantum capsule untuk memperkaya kematangan representasi fitur awal dan Quantum Vote Transformation untuk transformasi fitur dengan peningkatan representasi spasial. Prosedur routing dioptimalkan melalui algoritma hybrid quantum routing dengan mengintegrasikan metrik quantum similarity. Model dievaluasi menggunakan tiga dataset dengan tingkat kompleksitas berbeda, yaitu MNIST, rekaman CCTV, dan citra medis Pneumonia. Hasil eksperimen menunjukkan bahwa capaian akurasi tertinggi sebesar 0,9993 pada dataset MNIST. Dibandingkan dengan model baseline klasik, model berbasis kuantum mendapatkan peningkatan akurasi tertinggi sebesar 6,67% pada dataset rekaman CCTV.
====================================================================================================================================
Image classification is an important field of research for technological advancement and has practical applications in everyday life, such as accident detection systems and medical diagnosis. Capsule Networks are a type of deep learning model used for image classification, with the advantage of capturing spatial information through routing algorithms. However, this model has limitations due to its simple feature-extraction techniques and the high computational load of the routing process. Researchers have demonstrated that quantum computing can enhance the performance of deep learning models by leveraging the properties of quantum mechanics. In this study, we constructed quantum-based feature extraction and routing layers for the Capsule Network model. The feature extraction component was developed using a Variational Quantum Circuit with two stages: a quantum capsule to enhance the maturity of the initial feature representation, and a Quantum Vote Transformation to refine features and improve their spatial representation. The routing procedure was optimized using a hybrid quantum routing algorithm that integrates a quantum similarity metric. The model was evaluated using three datasets with varying levels of complexity: MNIST, CCTV footage, and medical images of pneumonia. The experimental results show that the highest accuracy, 0.9993, was achieved on the MNIST dataset. Compared to the classical baseline model, the quantum-based model achieved the highest accuracy improvement of 6.67% on the CCTV footage dataset.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Feature Extraction, Image Classification, Quantum Computing, Variational Quantum Circuit, Capsule Network, Ekstraksi Fitur, Klasifikasi Citra, Kompulasi Kuantum, Lapisan Sirkuit Kuantum |
| Subjects: | Q Science > QA Mathematics Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44002-(S3) PhD Thesis |
| Depositing User: | Ridho Nur Rohman Wijaya |
| Date Deposited: | 06 Aug 2026 06:08 |
| Last Modified: | 06 Aug 2026 06:08 |
| URI: | http://repository.its.ac.id/id/eprint/144154 |
Actions (login required)
 |
View Item |