NEDO-Commissioned Project JPNP23003 / Cyber-Physical Development of Quantum-Classical Hybrid Technology
News
Mission

Answering complex societal challenges with quantum × AI.

Addressing societal challenges such as maintaining and strengthening Japan's industrial competitiveness and achieving carbon neutrality requires the ability to process exponentially growing and increasingly complex computations in a timely manner. This means we need technology capable of solving, within a realistic time frame, problems that classical computers alone could not solve — whether in scale or in kind. Combining quantum computing technologies, including quantum-inspired approaches, with artificial intelligence (AI) is expected to dramatically improve computing power and enable more advanced use of data.

This project aims to accelerate the creation of applications that leverage quantum-AI hybrid technology. We are developing a common library that implements algorithms in a reusable form for a wide range of use cases, and building an ongoing management framework to support it.

Research Themes

Research Themes

From building the common library foundation to developing optimization and machine learning algorithms,
the following themes are being pursued in parallel.

Research Item 1-1

Research and Development of Common Library Usage Environment

AIST FY2023-2027

This R&D covers quantum-AI applications and the environment for using the common library, as well as clarifying the management framework needed to promote the library's adoption. Specifically, toward standardizing how the common library is used from quantum-AI applications, we develop a quantum-AI workflow orchestrator that handles execution-environment allocation and scheduling for tasks within quantum-AI applications, build out the quantum-AI software stack, and develop a common library usage environment with a user portal. We also clarify the management framework by building a "field" where the common library can be trialed. In addition, we research and develop the common library's external specifications — including the input/output data formats for programs built on quantum-AI hybrid algorithms, the library's configuration and required functions, and how data is exchanged between quantum and AI.

Research Item 1-2

Research and Development of External Specifications for the Common Library

Chodai, DEVEL FY2023-2027

Chodai Co., Ltd., together with the National Institute of Advanced Industrial Science and Technology (AIST), has published a tutorial*1 to support application development using cutting-edge "quantum-AI hybrid technology." This initiative aims to nurture the future IT talent needed to help solve increasingly complex societal challenges through information technology. The teaching materials include concrete sample programs, and the goal is to create an environment where anyone can verify and experience real behavior and performance using ABCI-Q, Japan's leading quantum-AI computing infrastructure operated by AIST.

Research Item 2

R&D of the Black-Box Optimization Common Library

Keio University FY2023-2025

This R&D extends FMA, a method for black-box optimization using hybrid algorithms that combine Ising machines — including quantum annealing machines — with AI. The conventional FMA only handles black-box optimization problems where variables are binary, there are no constraints, and there is a single objective function. We expand FMA's scope to support three or more discrete values, constraints, and multi-objective optimization, and develop new algorithms and a library that overcome the limitation of not being able to handle more data than the number of variables an Ising machine can hold.

Research Item 3

R&D of the Quantum Machine Learning Common Library

Tohoku University FY2023-2025

This R&D develops algorithms capable of finding exact solutions using quantum and simulated-quantum annealing machines. Specifically, we clarify the clustering algorithms and solution accuracy required for applications, and — building on the development of clustering algorithms using Ising machines — implement and evaluate algorithms that use quantum and simulated-quantum annealing machines, then turn them into a library.