- JST Home
- /
- Strategic Basic Research Programs
- /
- CREST
- /
- Research Director/
- Computational Foundation/
- [Computational Foundation] Year Started : 2018
Professor
Faculty of Science and Technology
Keio University
Toru Ishihara | Professor Graduate School of Informatics Nagoya University |
Katsuki Fujisawa | Professor Institute of Mathematics for Industry Kyushu University |
An efficient graph processing engine at edge devices is a key to realize a concept of Society 5.0 where various information obtained in the physical world is processed and optimized in a cyber world, and then feedback to the physical world is provided in realtime. In this research, we develop a highly efficient edge graph processing framework including a low-power and low-latency accelerator architecture, system LSI design, and its software environment through co-design with real graph applications. We also consider cooperation between graph processing and AI/annealing computation in the framework.
Professor
Graduate School of Information Science and Technology
Osaka University
Jun Tanida | Professor Graduate School of Information Science and Technology Osaka University |
Masanori Hashimoto | Professor Graduate School of Informatics Kyoto University |
This project aims to establish a new generation of computing technology with photonic neural networks by combining advanced neural and photonic computing technologies. From the viewpoint of spatiotemporal dynamics, this project develops recurrent neural network models for photonic implementation, and propose new computing principles and hardware implementation of photonic neural networks.
Professor
Institute of Innovative Research
Tokyo Institute of Technology
Akira Sakai | Professor Faculty of Science Hokkaido University |
Atsuyoshi Nakamura | Professor Graduate School of Information Science and Technology Hokkaido University |
Shinichi Minato | Professor Graduate School of Informatics Kyoto University |
Through inter-disciplinary collaboration among architecture, machine learning, discrete algorithm, and mathematical science, this project targets creating a new science-technology paradigm for realizing highly secure, dependable, and energy efficient “intelligent computing” that sustains Society 5.0 and beyond. Identifying energy minimization behind machine learning and combinatory optimization problems as a guiding principle, also spatio-temporal computing as an optimal architecture for drastically changing computation workloads, the project will establish a multi-purpose, self-learnable, spatio-temporal energy minimization HW/SW platform. Mathematical science and computer science will together ground break a new research domain.