Research outline
The application of new supercomputers to large-scale graph analysis and data processing is garnering attention. Applications of graph analysis include evacuation guidance plans for major disasters, and the effective use of large-scale data such as social networks for public policy and business management. However, the amount of computation, data volume, and power consumption required for these tasks are enormous, making traditional methods inadequate. Research is being conducted using high-performance computing technologies to enable ultra-large-scale graph processing. In the fiscal year 2015 (at the international conference ISC15, June 2015), we achieved the world's highest performance in Graph500 by solving large-scale graphs using our proprietary software, demonstrating the capabilities of supercomputers like Fugaku and the K computer in Japan. These supercomputers have achieved outstanding results in the Graph500 benchmark (8th, 10th to 18th, 20th to 25th editions).
Additionally, we are working on fast computation of optimization problems and their application in the real world. Semidefinite programming (SDP) problems, for example, have a wide range of applications in combinatorial optimization, systems and control, data science, financial engineering, and quantum chemistry, and are among the most notable optimization problems currently. We have succeeded in solving the world's largest SDP problems quickly and stably through the development of fast and stable algorithms and large-scale parallel computing on supercomputers. We are actively advancing research to solve complex and unsolved problems in the real world using mathematics and computers, in collaboration with other research institutions and companies. These achievements have been recognized with awards, such as the Minister of Education, Culture, Sports, Science and Technology Award in Science and Technology (Research Category) in 2017.
In recent years, efforts to realize a so-called super smart society (Society 5.0, etc.) through the combination and integration of the latest technologies for safety, security, and convenience are being promoted worldwide. Advances in ICT have made it possible to model phenomena occurring in the real world on computers in advance and to implement simulations and optimizations in response to environmental changes, thereby realizing Cyber-Physical Systems (CPS) as business models. Currently, in collaboration with many private companies, we are developing a CPS Mobility Optimization Engine (CPS-MOE), which uses massive sensor data (movement of people and objects, etc.) and open data (Wi-Fi movement history, etc.) to optimize, learn, and simulate in cyberspace, contributing to the creation of new industries, reduction of costs and waste, and calculation of optimal control schedules for transportation systems. Specifically, we are advancing the proposal and development of new mathematical and informational technologies for expressing, predicting, optimizing, and controlling the following three mobilities for the realization of CPS-MOE:
1. Mobility of information (people's interests, intentions): User clustering using web access movement data and users' latent interest levels.
2. Mobility of people and objects: Location detection and tracking (deep learning), congestion detection, flow optimization, and visualization.
3. Mobility of transportation (optimal driving): Route optimization, delivery optimization, MaaS (e.g., bike-sharing).
