Keynote & Tutorial Talks

Speaker: Prof. Koichi Shinoda (Tokyo Institute of Technology)
Title: Deep Learning and High-Performance Computing
Abstract: It has been over ten years since deep learning has emerged. Its performance has become equal to or better than that of human’s on several tasks including image classification and on speech
recognition. On the other hand, the computational costs for its development and implementation has become larger, which make it difficult to apply it to larger-scale tasks such as video recognition. To solve this problem, it is important to develop computer systems specified for deep learning. This talk will introduce our co-design approach for fast and cost-effective deep learning algorithm platform. In this project, researchers from machine learning and from high performance computing corporate to optimize algorithms and architectures simultaneously.

Speaker: msyksphinz (Handle name) (FPGA Development Diary Author)
Title: Changing semiconductor world around open-source CPU architecture “RISC-V”
Abstract: RISC-V has been attracted as an open source CPU architecture. RISC-V foundation are getting attension by attending many companies. Semiconductor industry, which had tended to close in the past, is following the open architecture by RISC-V. And, many semiconductor vendors that have had a closed policy changed their policy, and they started to open their IPs. While explaining the latest trends in open hardware and its ecosystem centered on RISC-V, I will introduce how the semiconductor development industry is changing.

Speaker: Takeo Imai (Idein inc. / National Institute of Informatics)
Title: Basics of deep learning compilers and their recent trends
Abstract: According to the widespread use of deep learning in recent years, deep learning compilers that optimize a deep neural network (or DNN) and generate executable code that runs on various hardware began to be used. This tutorial introduces the basics of deep learning compilers: Their history, basic structures, and recent technological trends.