Results of the FY2024 MITOU Target Program: Software Development Utilizing Quantum Computing Technology

In-page Links

Results of the FY2024 MITOU Target Program: Software Development Utilizing Quantum Computing Technology

Category 1: Software Development for Annealing Machines

PM: TANAKA Shu

Development of an automated shift allocation service with low implementation burden for individual tutoring schools utilizing quantum annealing

"Zyuku-Koma" is an automated scheduling service for private tutoring schools utilizing quantum annealing technology. It achieves high-precision scheduling, overcoming challenges faced by traditional manual and semi-automatic software. The system seamlessly handles tasks from schedule collection for timetable generation, employing 14 algorithms to optimize complex conditions such as student-teacher compatibility, subjects, consecutive sessions, and 1:N ratios. Features include OCR capabilities and Gantt chart displays, ensuring efficient UX. Designed for flexibility, it offers free and paid plans with immediate usability. This service contributes to DX advancement, operational efficiency, and alleviating labor shortages in the tutoring industry, serving as a model case for the societal implementation of quantum annealing technology.

• View the details of Development of an automated shift allocation service with low implementation burden for individual tutoring schools utilizing quantum annealing (in Japanese)

Creators: FUKUDA Shuichi

Development of a quantum annealing based training framework for Boltzmann machines

This project developed a learning framework for Boltzmann Machines using quantum annealing machines, aiming to create an efficient, low-power AI learning environment. Traditionally, utilizing quantum annealing required specialized knowledge, but the developed software simplifies its application by allowing users to specify key parameters. We implemented a sparse Boltzmann Machine aligned with the native physical topology of the quantum annealing hardware and successfully trained it on handwritten-digit data. Compared with conventional GPU-based approaches, this method significantly reduces power consumption, contributing to the realization of carbon-neutral computing environments. The proposed technology is expected to accelerate the adoption and research of quantum–classical hybrid techniques and to support further advances in AI technologies.

• View the details of Development of a quantum annealing based training framework for Boltzmann machines (in Japanese)

Creator: MATSUURA Takumi

PM: TANAHASHI Kotaro

Development of next-generation automatic routing tool utilizing annealing machines

This project developed a next-generation automated routing tool utilizing quantum annealing machines to enhance the efficiency of printed circuit board (PCB) design. Routing problems, considering factors such as length, via count, and cross-avoidance for electrical properties, were defined as combinatorial optimization problems, and algorithms were constructed. At the global routing stage, paths were generated using Delaunay triangulation and the A* algorithm, followed by optimization via quantum annealing. Validation tests successfully produced feasible solutions for small to medium-sized PCBs within short timeframes. For large-scale PCBs, a dynamic candidate path updating method was implemented, effectively reducing constraint violations. This tool is expected to reduce design time, strengthen the supply chain of electronic devices, and offer broad applicability across various industries.

• View the details of Development of next-generation automatic routing tool utilizing annealing machines (in Japanese)

Creators: KATO Shunsuke, NAGAYAMA Kotaka, TOYAMA Kota

Development of a surface modeling tool for alloy catalysts using an annealing machine and user support documentation

"SurfQit" is a tool that leverages quantum annealing technology to make alloy catalyst surface modeling efficiently, aiming to support the realization of carbon neutrality by 2050. The software optimizes metal arrangements on catalyst surfaces and constructs models based on formation energy, significantly reducing computational costs compared to traditional methods. It provides functionalities such as predicting metal configurations for binary and multi-component alloys, analyzing and visualizing structures, and enabling users to add required training data. Additionally, user support documents and a browser-based environment for experiencing the software have been developed to promote adoption among researchers worldwide. This technology accelerates catalyst development, facilitates various catalysis such as conversion of carbon dioxide into fuel and hydrogen energy utilization, and contributes to building a sustainable society.

• View the details of Development of a surface modeling tool for alloy catalysts using an annealing machine and user support documentation (in Japanese)

Creators: SAMPEI Hiroshi, SAEGUSA Koki

PM: TAMURA Ryo

Development of a support engine for autonomous robots for disaster relief

This project developed a motion planning method for an autonomous robotic arm intended for use in disaster recovery environments. By leveraging quantum annealing, we designed a new algorithm that enables a multi-joint robotic arm to work efficiently and safely in complex, cluttered environments. The approach involved three stages: terminal trajectory search, posture difference exploration, and motion adjustment, facilitating real-time action planning. Additionally, a simulation environment was created on Unity to quantitatively compare various planning methods. This technology is expected to enhance investigation and operational efficiency in disaster sites and nuclear disaster recovery areas, preventing secondary disasters and reducing radiation exposure risks.

• View the details of Development of a support engine for autonomous robots for disaster relief (in Japanese)

Creators: FUJIMOTO Hiroki, SAKAGUCHI Ryo

Development of computational support middleware for annealing machines using machine learning-based dimensionality reduction and QUBO formulation

This project developed a novel optimization method using an autoencoder to enhance the efficiency of quantum annealing machines. Conventional One-hot encoding increased the number of bits with variable binarization and led to frequent infeasible solutions. By leveraging autoencoder, the new method reduced the number of bits and increased the occurrence rate of feasible solutions. Evaluation using the traveling salesman problem demonstrated high accuracy in obtaining optimal solutions. This method is expected to accelerate the development of optimization technology and promote the widespread adoption of quantum annealing and quantum computing technologies by providing a new approach to complex combinatorial optimization problems.

• View the details of Development of computational support middleware for annealing machines using machine learning-based dimensionality reduction and QUBO formulation (in Japanese)

Creators: YAMASHITA Masashi, ABE Tetsuro

Category 2: Software Development for Gate-based Quantum Computers

PM: TOKUNAGA Yuki

Development of a quantum circuit optimization method using ZX-Calculus and reinforcement learning

This project aimed to enhance quantum circuit optimization technology by developing "YN-Optimizer," which utilizes reinforcement learning. Based on ZX-Calculus, this method transforms quantum circuits into graphs and dynamically optimizes simplification operations, overcoming the limitations of traditional fixed-order optimization approaches. Key technologies include the C++-based "CZX" library for speed enhancements, reinforcement learning-based "YN-RL," and Monte Carlo tree search integrated "YN-MCTS," which improved gate reduction rates in quantum circuits. These achievements are expected to boost quantum computation accuracy and feasibility, highlighting the importance of this foundational technology for the practical implementation of quantum computing. Comparative analysis with non-ZX-Calculus optimization methods confirmed competitive performance.

• View the details of Development of a quantum circuit optimization method using ZX-Calculus and reinforcement learning (in Japanese)

Creator: NOGAMI Hiroki, YOSHIOKA Malick

Development of a quantum program creation tool that utilizes 3D CG and a game controller for an intuitive user experience

"CoQtail" is an integrated development environment designed to facilitate entry into quantum computing technology. This tool enables intuitive creation of quantum algorithms and quantum error correction codes, while being optimized for usage of game controllers. It offers functionalities such as quantum circuit editing, quantum state visualization, and programming efficiency improvements through block-based features, catering to users ranging from beginners to experts. Additionally, its 3DCG-based error correction code visualization and creation capabilities support a wider variety of codes compared to traditional tools. This technology is expected to accelerate the realization of FTQC (Fault-Tolerant Quantum Computers) and contribute to the practical application of quantum computing by increasing new entrants and enhancing development efficiency. The project plans to continue expanding its features to create an even more user-friendly environment.

• View the details of Development of a quantum program creation tool that utilizes 3D CG and a game controller for an intuitive user experience (in Japanese)

Creators: FILHO Nilton F. G., KOJI Shinya, BUCEK Rodrigo A.

PM: FUJII Keisuke

Development of quantum feature mapping methods applicable to real-world problems

This project aims to streamline and advance quantum machine learning through the development of the "QXMT" (Quantum eXperiment Management Tool) and an automated quantum circuit generation system. QXMT assigns unique IDs to each experiment and standardizes the management of datasets, devices, and evaluation methods to ensure reproducibility. Designed for users ranging from beginners to experts, it supports the design of quantum feature maps and provides automated evaluation metrics. Additionally, a system leveraging LLM (Large Language Models) was developed to automate the generation and improvement of quantum circuits. The resulting quantum feature maps demonstrated comparable accuracy and versatility to existing methods. This system is expected to accelerate research and development in quantum machine learning and promote its practical application in industrial and academic fields.

• View the details of Development of quantum feature mapping methods applicable to real-world problems (in Japanese)

Creators: SAKKA Kenya

Development of new features for quantum programming contest "QCoder"

"QCoder" is a platform aimed at promoting quantum technology and nurturing quantum talent through quantum programming contests. This project introduced new features, including the "Writer System," allowing users to propose contest problems; the "Discussion Feature," facilitating user interactions; and the "Announcement Feature," sharing updates on upcoming contests. Furthermore, the release of an English version significantly increased international users' participation, establishing a global user base. Multiple contests were held, offering diverse problem sets that balanced competitiveness with educational value. These efforts not only enhance interest in quantum technology but also serve as a foundation for advancing quantum computing and its societal implementation.

• View the details of Development of new features for quantum programming contest "QCoder" (in Japanese)

Creator: YUKIYOSHI Kein, TAKATA Shunya

PM: YAMAMOTO Naoki

Development of continuous-variable optical quantum reservoir computing technology for quantum information processing

This project developed continuous-variable optical quantum reservoir computing technology to advance optical quantum computing and quantum information processing. The system enables direct input of quantum states into reservoir computers, surpassing the performance of classical reservoir computing. Specifically, the success probability of quasi-Bell measurements was improved compared to traditional methods, highlighting potential applications in quantum computing and quantum communication. For instance, it contributes to quantum computing based on cat codes and addressing challenges in long-distance quantum communication. These achievements lower barriers to the practical implementation of quantum computing, reduce quantum communication costs, and are expected to contribute to building a sustainable society.

• View the details of Development of continuous-variable optical quantum reservoir computing technology for quantum information processing (in Japanese)

Creator: KOJIMA Souta, KIRYU Shohei, SHIMIZU Tadaomi

Development of methods and applications for solving practical machine learning problems on small-scale hardware

This project developed the "LeanQML" algorithm to enable practical-level problem-solving using small-scale quantum hardware and implemented it as a Web API. It offers quantum ensemble computing and quantum reservoir computing algorithms, providing an accessible environment for users without prior quantum computing knowledge. Compared to classical machine learning methods such as LightGBM, the tool achieved comparable or superior performance. By lowering barriers to entry for quantum technology, this tool facilitates the expansion of research and development (R&D) and proof-of-concept (PoC) initiatives utilizing quantum machine learning. Additionally, it addresses the shortage of engineers by enabling AI engineers to easily experiment with quantum algorithms, accelerating the societal implementation of quantum computing.

• View the details of Development of methods and applications for solving practical machine learning problems on small-scale hardware (in Japanese)

Creator: TAKEDA Naoyuki, AGO Taichi, KANEKO Kazuya