Keynote Talks

Chair: Michihiro Koibuchi (NII), Shoichi Hirasawa (NII)

Speaker: Naoto Fukumoto, Fujitsu

Title: Performance tuning of applications for large-scale computer systems

Abstract: It is difficult to achieve high performance on large-scale computers. Various techniques and know-how are required, but this kind of information is often not disclosed. In this talk, the speaker will give an overview of tuning for large-scale computers, and introduce examples of tuning for Linpack benchmark and deep learning on famous Japanese supercomputers.

Chair: Keiichiro Fukazawa (Kyoto Univ.), Hiroyuki Takizawa (Tohoku Univ.)

Speaker: Yohei Miyake (Kobe University)

Title: A Decade of Effort in HPC toward Realistic Scale Spacecraft-Environment Interaction Simulations

Abstract： Rapid progress of space science and engineering over the last decades is greatly attributed to advanced use of computer simulations. However, some of numerical models, e.g., particle-in-cell plasma simulations, have faced significant difficulties in exploiting computational power of modern HPC systems. We, working in the field of space science, have implemented a long- standing research collaboration with experts in the computer science community. The talk reviews the effort over this period to optimize particle- based plasma simulation model for modern and changing HPC systems. We also introduce our recent activities to develop a multi-scale framework for further realistic space simulations.

Chair: Sayaka Kamei (Hiroshima University), Fukuhito Ooshita (NAIST)

Speaker: Stéphane Devismes, Université de Picardie Jules Verne

Title：Self-Stabilizing Leader Election in Highly Dynamic Networks

Abstract: Modern networks (e.g., MANET, VANET, and DTN) are prone to both faults and frequent alteration of their topology (i.e., the addition or removal of links). Self-stabilization is a versatile fault-tolerant property. In this talk, I will present recent advances in self-stabilization for highly dynamic systems. I will focus on the fundamental problem of leader election.

Chair: Hideharu Amano, Nobuhiko Nakano

Speaker: Shaswot Shresthamali (Keio University)

Title: Reinforcement Learning for Energy Harvesting Wireless Sensor Nodes

Abstract: Energy Harvesting Wireless Sensor Nodes (EHWSNs) have become very popular as edge devices in the Internet of Things (IoT) ecosystem. Since they harvest energy from their working environment, they can operate perpetually by maintaining energy neutrality. This makes them very attractive as long-term sustainable solutions for IoT. In this talk we will investigate how EHWSNs can leverage Reinforcement Learning (RL) to learn intelligent energy management policies that can optimize and adapt even when the working environment is complex and unpredictable. However, RL requires significant amounts of time and computation in addition to the risks associated with trial-and-error learning. We will discuss why the learning process is so inefficient and what solutions are feasible in the context of EHWSNs. Specifically we will look at how clever problem formulation and distributed learning can help overcome some of the challenges in implementing RL for EHWSNs.

Chair: Toru Nakanishi (Hiroshima University)

Speaker: Kazumasa Omote, University of Tsukuba

Title: Security Risks for Blockchain and Cryptoasset

Abstract: The blockchain satisfies not only the cryptographic aspect of tamper resistance, but also the network aspect of high availability, and furthermore, it can directly handle money, which is called a cryptoasset. On the other hand, it has been observed that attacks have actually been carried out using the properties of tamper resistance and high availability. Because of the direct handling of money, attacks on blockchains are very active. In this talk, we present recent researches on security risks related to blockchain and cryptoassets, including blockchain poisoning attacks, unauthorized access to cryptoasset nodes, and the risk of worthless cryptoassets.