Automated generation of quantum computing models using deep learning

Quantum computing offers high power with low energy compared to classical computing, demonstrating proven performance and significant potential. Despite extensive research to leverage these advantages, obstacles exist in both hardware and software for their active use. The thesis study is dedicated to exploring ways to harness the full potential and advantages presented by the quantum computing paradigm. The research delves into various aspects, including the properties of superposition and entanglement, methods for encoding classical data into the quantum environment, the development of quantum computing simulators, creation techniques for quantum circuits, optimization methods for both quantum circuits and code, and the conversion of quantum unitary matrices into quantum circuits and code. The main scientific contributions obtained as a result of the research and investigations include an efficient data coding method that uses the superposition feature to transfer classical data to the quantum environment, an efficient and scalable quantum circuit design for deep reinforcement learning, an efficient quantum computing simulator that will facilitate quantum circuit and code design, It has been provided in the areas of methods that create quantum circuit and code data sets and development of methods that create deep learning-based automatic quantum computation models. The proposed efficient and scalable circuit design for deep reinforcement learning has demonstrated commendable success rates even with a low number of qubits. Furthermore, in contrast to alternative methods, the number of qubits in the circuit ensures the number of elements in the action set, irrespective of the number of parameters. The proposed quantum computing simulator has achieved an impressive efficiency of 100% in both the design and code stages, outperforming other simulators in the comparison. The method for generating a quantum circuit and code dataset has achieved a 100% accuracy and success rate in producing truth tables for quantum circuits, TFC codes, quantum unitary matrices, and quantum circuit images, tailored to the specified number of qubits and depth. In the method for creating a deep learning-based automatic quantum computation model, a notable success rate of 94% was attained through the deep learning model that converts the truth table into a quantum unitary matrix, coupled with the system that transforms the unitary matrix into a quantum computation model. Consequently, various recommendations have been proposed across multiple domains to harness the potential and advantages of quantum computing. The thesis study has demonstrated the feasibility of converting classical circuits, truth tables, and algorithms used in classical computations into their quantum counterparts.