Undergraduate Thesis: Quantum Image Classifier Design with Data Re-uploading Quantum Convolution and Data Re-uploading Classifier Scheme
A repository for finishing my undergraduate thesis titled:
Quantum Image Classifier Design with Data Re-uploading Quantum Convolution and Data Re-uploading Classifier Scheme.
Advisors: Prof. Andriyan Bayu Suksmono M.T., Ph.D. and Ir. Nugraha, Ph.D.
A flow diagram of the Data Re-uploading Quantum Convolution scheme.
The need for computational power keeps increasing as industry and academia's problems get harder and harder to solve. Applications such as simulation of large quantum systems like molecules or solving large linear systems can be very expensive in computational cost. This has become one of the reasons for quantum computing development, a computational method that employs characteristics and theories of quantum systems for information processing. Quantum computers promise us exponential speed-up for these kinds of problems.
Although quantum computers' development has been growing rapidly in recent years, the theoretical and technological challenges remain a barrier for a large-scale quantum computer. Quantum computers that exist today have severe limitations, such as limited qubits and limited gate operations due to noise in the processes. Variational Quantum Algorithms (VQA) have arisen to be one of the promising strategies in dealing with these limitations. Applications across the fields that employ this strategy have been proposed, including image classification as quantum machine learning applications.
This research proposed a modification scheme of the VQA-based Data Reuploading Classifier (DRC) for MNIST classification. A binary, four-class, and eight-class classification task reached 99.7%, 96.5%, and 86.25% of testing accuracy, respectively, using the Principal Component Analysis (PCA) for dimensionality reduction and Data Re-uploading Classifier with binary representation (DRC-BR) for classification. An improvement in accuracy compared to the previous related VQA works. This research also proposed a DRCbased quantum convolution scheme. Without using PCA, quantum convolution with DRC-BR classifier for binary and four-class classification task achieved 98.9% and 89.5% of testing accuracy, respectively. A respectable result compared to the classical Convolutional Neural Network with the same amount of parameters.
Note: all contents here are written in Indonesian as required by the university. If you are non-Indonesian but interested in this research and would like to receive a summary in English, please feel free to contact me.
This thesis project is supported by grants from Radio Telecommunication and Microwave Laboratory under Prof. Andriyan's supervision. Starting from January 2021, this thesis project's affiliation changed to the Quantum Technology Laboratory, a newly founded laboratory at Bandung Institute of Technology.