Members of Togawa Laboratory
Prof. Nozomu Togawa
Professor, Department of Computer Science and Communications Engineering, School of Fundamental Science and
Engineering, Waseda University (Back row,
3rd from left)
My research areas are integrated circuit design (IC chip design technology), hardware security, and a field of electrical engineering and computer science. Our group conducts research on "application and system design" based on computer science and engineering and information and communication sciences. Here, hardware, software, and services are integrated, with keywords such as wearable devices, augmented reality (AR), geography and location information, machine learning and deep learning, information security, IoT, environmental power generation, and quantum computing. In particular, we are actively engaged in "joint research between industry, government, and academia", which is a distinctive feature of our research.
I started my research activities in the field of "integrated circuit design". Large-scale combinatorial optimization operations," which are now a trend, were developed in the field of integrated circuit design. In order to create high-performance integrated circuits, there has been a constant need since the 1990s to solve optimization problems by computers, such as optimal placement of elements and gates, graph-based combinatorial optimization, and algorithm development.
The field of integrated circuit design research can be divided into several layers.
I have been working in two of these layers: "system design using semiconductor devices (especially in the area of FPGAs)" and "computer-aided design or electronic design automation (EDA)".
What triggered my involvement in the quantum computing field was a joint research project on semiconductors with major Japanese electronics manufacturers. At that time, these Japanese electronics manufacturers had already recognized quantum annealing as a breakthrough technology for high-speed computation of combinatorial optimization problems; the race to develop Ising machines began simultaneously. In the course of our joint research, I began to explore the application of Ising machines and to research and develop implementation techniques.
The major Japanese electronics manufacturers involved in Ising machine development were originally companies that researched, developed, manufactured, and sold domestically produced personal computers. There were many engineers within these companies who excelled in hardware development, and many of them must have found a way to take advantage of the new technology and new field of quantum computers. We feel the strong desire of these manufacturers to create products and businesses that outperform the rest of the world.
Quantum computers use physical methods to manipulate quanta as computational elements, and quantum computer researchers tend to be physics experts. In this community, there are few researchers from the EECS fields like myself. I have had an interest from the beginning in exploring applications for quantum computers and developing applications. Without a background in physics, it is difficult to understand quantum computation theory, and the hurdles to utilizing quantum computers would be. With Fixstars Amplify, even general information technology engineers can easily develop applications.
With convenient tools and development environments such as Fixstars Amplify, I believe that the future of quantum computer utilization technology will see more integration with different fields. For example, researchers from different fields have already begun to work together to conduct joint research on quantum computer technology, and the influence of their research results is expanding.
I have conducted various joint research projects with private companies, other universities, and research institutes using Fixstars Amplify. Here are some typical examples.
This is a national project which began in 2018 and involves many private companies, universities, and research institutes, including Fixstars Corporation. In this joint research, Waseda University was the representative of this project and I was the R&D leader. The ultimate goal was to utilize the high-speed computation to solve "combinatorial optimization problems" in various industries such as automobiles, logistics, finance, drug discovery, and factory production management. However, at present, the high degree of difficulty in utilizing the Ising machine was a major barrier, and in order to reduce this difficulty, we decided to first study a "common software platform for Ising machines.” Fixstars Amplify is being used as a common software platform to abstract a hardware layer of Ising machines developed by domestic and overseas companies and make them easier to use.
New Energy and Industrial Technology Development Organization (NEDO)I was the principal researcher and conducted joint research with Fixstars Corporation. This initiative is to research and develop the infrastructure for societal implementation of a system that highly integrates real and virtual space through next-generation accelerators to solve relevant problems in all industries. Currently, various next-generation accelerators have been proposed. We have built a system architecture that allocates optimal computing resources according to the characteristics of these accelerators. In order to cope with the growing amount of data and complex computational processes resulting from the IoT and DX, conventional computers will no longer be able to handle the growing amount of data and complex computational processes. In recent years, research and development of "next-generation accelerators (quantum annealing, Ising machines, NISQ devices, and error-tolerant quantum computers)" have progressed. However, next-generation accelerators are not universally applicable to all problems, and are characterized by their suitability or unsuitability depending on the scale and characteristics of the problem. In fact, there are situations in which a conventional computer is more appropriate than a next-generation accelerator. In other words, it is important to use different computing resources for different situations. For this reason, this research introduced the concept of "co-design" (a mechanism for optimal allocation of computing resources), which includes the hybrid use of next-generation and conventional accelerators.
Furthermore, the research results of the SIP were utilized in a logistics warehouse, and the system went into actual operation in October 2022 as an automatic system*6 for optimal allocation of personnel. In addition to the optimization of truck delivery routes, which will be discussed later, the optimization of personnel allocation in warehouses is also considered an important means of contributing to solving issues in the logistics industry, and is a theme of great social significance and strong need.
Quantum Science and Technology AgencyResearch is being conducted with the goal of commercializing an efficient and optimized delivery planning system by 2025. Calculating truck delivery routes is extremely difficult with conventional computers and requires a long calculation time as the scale of the problem increases, but Fixstars Amplify is expected to have the potential to easily calculate these routes in a short timespan. The logistics industry continues to grow with the expansion of internet services and mail-order businesses, but at the same time, it has long faced issues of human resource shortages and overwork. In order to solve this problem, the optimization of truck delivery routes is now attracting a great deal of attention.
Undergraduate and graduate students in my group use Fixstars Amplify.
Many of the students have only acquired programming skills in their courses and have not necessarily studied quantum mechanics or quantum computers beforehand. Many students study combinatorial optimization problems for the first time after being assigned to the group. They gradually learn how to use Fixstars Amplify with their tutors in their doctoral courses or by operating it on their own while referring to the demos and documents*8 on the Fixstars Amplify website.
Often, the students needed to run their combinatorial optimization programs all day long on Fixstars Amplify Annealing Engine (AE). However, because there is no pay-as-you-go system for any plan, I did not have to worry about costs, time limits, capacity limits, or other restrictions.
When using an annealing machine, it requires a high level of expertise in informatics and physics in general. With Fixstars Amplify, however, the amount of expertise and work required is greatly reduced because Fixstars Amplify handles many parts of the process to use an annealing machine automatically. For example, research often involves trial and error, where the formulation of the problem and the execution of the machine are repeated many times. Fixstars Amplify makes it possible to efficiently perform advanced formulations and easily set parameters to maximize the performance of the machine. As a researcher, I really appreciate the fact that we can streamline our research in this way.
Furthermore, we feel that Fixstars Amplify AE is most practical because of the large problem size of QUBO that can be solved and the speed and accuracy of the quadrature solution. Another attractive feature of Fixstars Amplify AE is its interface for retrieving information useful for research. This includes the display of multiple solutions, along with the history of updates to the best solution in the solution search.
Fixstars Amplify is very helpful because it supports multiple machines, allowing us to utilize the same combinatorial optimization program, but switch only the machines to be used.
Thus, when using Fixstars Amplify, the basic operation is relatively simple and the required programming skills are not astronomical, so you can quickly become proficient in a short period of time. I think Fixstars Amplify is attractive for many researchers since it is easy to incorporate into not only long-term R&D projects, but also into basic studies such as student classes and graduation theses, etc. I believe that cross-disciplinary research and studies using Fixstars Amplify will be further advanced in the future. I will continue to work with Fixstars Amplify in research and education.
I would like to continue to use Fixstars Amplify in research and education to expand the possibilities of application development using quantum computer technology in various industries and to contribute to solving real-world problems.
Through the results of the NEDO research project "Research and Development of a Common Software Platform for Ising Machines "*4, Fixstars Amplify has grown into middleware that is recognized as the de facto standard for the use of quantum computer technology in Japan. Many domestic Ising machines, leading overseas quantum computers, and mathematical optimization solvers can now be easily used with Fixstars Amplify, and the software is currently being used in a variety of applications. It will be commonplace for vendors developing Ising machines and quantum computers to prepare drivers for Fixstars Amplify in the future. Currently, many vendors provide SDKs specific to their own machines, but in order to accelerate the development of quantum computers by attracting a broad user base from various industries, there should be a trend toward the need for middleware that connects and integrates all quantum computers around the world. I believe that Fixstars Amplify will play a key role in this process.
*1 Combinatorial optimization problem
Combinatorial optimization problems are highly computationally demanding in various industries such as
manufacturing, logistics, and finance. It searches for the optimal combination for an objective among possible
combination patterns. For example, if we can calculate the optimal pattern for logistics transportation
planning or manufacturing planning, it will be useful for improving the efficiency of business activities and
predicting return on investment.
On the other hand, the more variables and conditions to be considered, the more rapidly the number of
combination patterns increases. As variables such as the number of vehicles and delivery destinations in truck
logistics, and constraints such as the amount of load capacity and availability of drivers, increase, the
calculation to find the optimal pattern becomes more difficult.
Examples: traveling salesman problem, production planning optimization, work shift optimization, route
optimization, etc.
*2 Ising Machine
An Ising machine is a computer specialized for solving approximate combinatorial optimization problems, also
known as an annealing machine.
Many combinatorial optimization problems can be converted into ground state search problems for the Ising
model, which is a mathematical model of magnetic materials. An Ising machine is a computer that finds an
approximate solution to this search problem.
And Ising machines are divided into two types. They are classical annealing machines (utilizing classical
computers and semiconductor CMOS integrated circuits) and quantum annealing machines (utilizing the principle
of quantum annealing).
*3 Conference presentation (Japan Society of Applied Physics): Structural Optimization of
Photonic Crystal Lasers by Quantum Annealing
https://www.waseda.jp/top/news/83121 (Note: Japanese only)
Presented at: The 83rd Japan Society of Applied Physics Autumn Meeting Lecture Number: 21a-A101-11 Date of
Publication: 2022/9/22
Author: Takuya Inoue, Yuya Seki, So Tanaka, Nozomu Togawa, Kenji Ishizaki, Susumu Noda
We applied quantum annealing to the structural optimization problem of a "photonic crystal laser," which is
expected to be widely used as a laser source for smart processing in the future, and obtained a device
structure with a nontrivial spatial distribution that is different from conventional designs. The obtained
structure was confirmed by calculations to have improved performance in all aspects compared to the
conventional design.
*4 NEDO (New Energy and Industrial Technology Development Organization): Research
Project
Development of AI Chip and Next-generation Computing Technology Enabling High-efficiency and High-speed
Processing>Research and Development Item 2: Development of Next-generation Computing Technology
https://www.nedo.go.jp/koubo/IT3_100063.html (Note: Japanese only)
*5 Strategic Innovation Program (SIP): Research projects
Optoelectronic information processing > 1. Research and development on next-generation accelerator
infrastructure > Adopted research theme: Research and development on co-design infrastructure integrating
next-generation accelerators
https://www.qst.go.jp/site/sip/35680.html
(Note: Japanese only)
*6 Sumitomo Corporation Group: Automatic Personnel Optimization System in Operation at
Belle Maison Logistico's Logistics Warehouse (Application of SIP Research Results)
Fixstars' Case Study (Belle Maison Logisco): https://www.fixstars.com/ja/services/cases/amplify-bellemaison (Note: Japanese only)
*7 Sumitomo Electric Industries press release: Optimization of truck delivery routes for
logistics business
https://sumitomoelectric.com/jp/sites/japan/files/2022-09/download_documents/prs110_0.pdf
(Note: Japanese only)
*8 Fixstars Amplify demo & tutorial, documentation
Demos & Tutorials
Documentation
* All information in this article is based on information available at the time of the interview (April 2023)
Dr. Togawa is researching the use of quantum computers and Ising machines from the standpoint of information engineering. We have heard that many "live" issues, including topics not introduced in this article, are being gathered by Dr. Togawa through many industry-academia collaborative projects. The hurdles to solving these issues are difficult, but I felt that clearing these hurdles will be the touchstone for whether or not next-generation technologies will be truly useful. As you can see from this interview, the Togawa laboratory is a heavy user and a big fan of Fixstars Amplify. We will continue to expand our services to support their diverse research activities.
Interviewer: Takuji Hiraoka (President and CEO, Fixstars Amplify Corporation)