Abstract

Currently, in the development of Self-driving vehicles, learning of images is implemented in GPGPU as the main constituent element in the server. However, since a large amount of data transfer occurs between the GPU and the memory, there is a doubt about the future in terms of performance improvement and power consumption. Furthermore, data flow computing, which is compatible with high performance and energy efficiency required for machine learning, is a promising candidate for GPGPU next generation, but several architectures have been proposed for practical use. Therefore, we consider data flow computing utilizing FPGA to quickly incorporate new technology useful for machine learning.

Biography

Yoshifumi Sakamoto is an associate partner of IBM Japan Ltd., and in charge of principal consultant of Advanced Information Technology for automotive. I received the Ph.D. degrees in computer science from Kyushu University, Fukuoka, Japan, in 2014. I was engaged in research of Systems development method utilizing Reverse modeling and Model-based Simulation in the graduate school of Kyushu University.

I joined IBM in 1985. I was responsible for the design and development personal computers, embedded systems and ASIC/SoC. After that, I was responsible for the project manager of ASIC and SoC development project. I responsible for the program manager of the multiple SoC development, and became a technical consultant to apply an advanced information technology into automobile. Currently, I am working as a consulting engineer of a vehicle centric control system and In-Vehicle Infotainment system.