Curriculum Vitae

2020-02-21

PDF version: Curriculum_Vitae.pdf

Hiroki Kanezashi is a post-doctoral researcher at the AIST-Tokyo Tech Real World Big-Data Computation Open Innovation Laboratory (RWBC-OIL) after finishing his Ph.D. at the Graduate School of Information Science and Engineering of the Tokyo Institute of Technology in June 2019.

His research field is large-scale data processing in high-performance computing (HPC) environments. His current research interests are graph analytics, including community detection and graph pattern matching, machine learning, and data analytics algorithms for financial services such as anti-money laundering (AML). He is supervised by Prof. Satoshi Matsuoka at Tokyo Institute of Technology and Prof. Toyotaro Suzumura in IBM T.J. Watson Research Center. He received the Best Paper Award on the 25th IEEE International Conference on High-Performance Computing, Data, and Analytics (HiPC 2018).

Education

  • 2015-2019 Ph.D. Mathematical and Computing Sciences, Tokyo Institute of Technology, Tokyo
  • 2013-2015 M.Sc Computer Science, Tokyo Institute of Technology, Tokyo
  • 2009-2013 B.Sc. Tokyo Institute of Technology, Tokyo

Master thesis

  • Title: Performance Optimization of Large-Scale Traffic Simulation on Parallel and Distributed Systems
  • Supervisors: Prof. Takehiro Tokuda and Prof. Satoshi Matsuoka
  • Description: It is indispensable to make full use of parallel and distributed systems with increasing demands for large-scale traffic simulation, but problems remain about insufficient scalability due to costs of synchronization by load unbalancing among compute nodes. To tackle this problem, we propose performance optimization methods for traffic simulations applying graph contraction to underlying road networks as well as introducing adaptive synchronization interval based on time- series traffic congestion. By applying these optimizations and running the simulation of the real-world Dublin city on 16 compute nodes of TSUBAME 2.5, the simulation performance has improved by 4 times with the proposed graph contraction method and improved by 3 times with adaptive synchronization method with comparison to regular 1 synchronization per step while keeping the simulation precision up to 10% difference.

Experience

Visiting Ph.D. Students

  • 2014-2014 Visiting Ph.D. Students, University Collage Dublin, Ireland, Validation of traffic simulation in Dublin city with open data, supervised by Prof. Toyotaro Suzumura

Research Assistant

  • 2011-2015 Research Assistant, Tokyo Institute of Technology, Japan, Research assistant about large-scale social simulation base

Teaching Assistant

  • 2015-2015 Tokyo Institute of Technology, Japan, Teaching assistant in Graduate School of Mathematical and Computing Science, supervised by Prof. Satoshi Matsuoka
  • 2015-2015 Tokyo Institute of Technology, Japan, Teaching assistant in Undergraduate School of Information Science, supervised by Prof. Satoshi Matsuoka

Visiting IBM T.J. Watson Research Center

  • 2016-2018 Visiting IBM T.J. Watson Research Center, IBM T.J. Watson Research Center, United States, Proposed an incremental community detection algorithm for large scale-free network, and as a future work, proposes more scalable graph data distribution methods for graph databases, supervised by Prof. Toyotaro Suzumura.

Languages

  • Japanese: Native
  • English: Intermediate

Computer skills

  • Languages: C, C++, Java, Python, X10
  • HPC Libraries: OpenMP, MPI, CUDA
  • Distributed: Hadoop, Spark
  • Graph Database: JanusGraph, Titan, Apache TinkerPop
  • Linux Server: CentOS 6 and 7

Supervisors

Satoshi Matsuoka

  • Professor
  • RIKEN Center for Computational Science (R-CCS) / Tokyo Institute of Technology (Department of Mathematical and Computing Sciences)

Toyotaro Suzumura

  • Research Staff Member
  • IBM T.J. Watson Research Center, New York, USA

Publications

BibTeX text: bibtex.txt

  • Hiroki Kanezashi and Toyotaro Suzumura. Performance optimization for agent-based traffic simulation by dynamic agent assignment. In Proceedings of the WinterSimulation Conference, 2015., 2015.
  • Hiroki Kanezashi and Toyotaro Suzumura. An incremental local-first commu-nity detection method for dynamic graphs. In Big Data (Big Data), 2016 IEEEInternational Conference on, pages 3318–3325. IEEE, 2016
  • Hiroki Kanezashi, Toyotaro Suzumura, Dario Garcia-Gasulla, Min-Hwan, and Satoshi Matsuoka. “Adaptive Pattern Matching with Reinforcement Learning for Dynamic Graphs.” 2018 IEEE 25th International Conference on High Performance Computing (HiPC). IEEE, 2018.