• Home
  • News
  • [Matsuo Lab., The University of Tokyo] Activity Report for 2019
  • [Matsuo Lab., The University of Tokyo] Activity Report for 2019

    Hello, this is Shimizu from Public Relations.

    This year marks the beginning of a new era, replacing the Heisei Era with the 2025 Era.
    The Matsuo Institute has also seen a flurry of good news over the past year, as Matsuo and other members have become more active in their respective fields.

    So, as a review of this year’s events, we would like to reintroduce some of the topics that have been discussed both inside and outside of Matsuo Lab.

    1. New member joins

    The Matsuo Lab has now grown to 68 members, and a new, reliable member has joined the group!

    Project Researcher Dr. Hiroshi Yamakawa

    Profile
    1987: B.S. in Physics, Tokyo University of Science
    1989: M.S. in Physics, Graduate School of Science, The University of Tokyo
    1992: Ph.D. in Electronic Engineering, Graduate School of Engineering, The University of Tokyo
    1992-2014: Fujitsu Laboratories Ltd.
    1994: Participated in the Real World Computing Project of MITI from the company
    2014-19: Director, Artificial Intelligence Laboratory, Dwango Inc.
    2015: Appointed as Visiting Researcher, Artificial Intelligence Research Center, AIST
    2015: Appointed as Representative, Whole Brain Architecture Initiative, a non-profit organization
    2015: Visiting Professor, Graduate School of Information Systems, The University of Electro-Communications
    2016: Appointed Senior Research Fellow, Keio Research Institute at SFC, Keio University
    2016-18: Editor-in-Chief, Japanese Society for Artificial Intelligence
    2017: Appointed as Visiting Researcher, School of Medicine, The University of Tokyo
    2018: Appointed as Visiting Researcher, RIKEN Center for Integrated Research on Innovative Intelligence
    2019: Appointed as Visiting Principal Investigator, RIKEN Center for Frontier Biosciences

    Comments
    “At the Matsuo Lab, I am involved in the research topic of world models. This is also called, for example, a mental model, in which an intelligent system, such as a human, accumulates experiences of interaction with the external world and models the surrounding environment independently of any specific goal or task. For a long time, the only way to learn a specific task solution was to provide direct supervised data. Beyond the development of deep learning, however, that situation has finally changed.
    It may be that it took 60 years for AI to begin to understand and solve tasks directly, rather than simply following a how-to.
    When you think about it, modeling the world is also the work of science, and this will greatly expand the possibilities for AI in the future. I myself was studying the cognitive association module for learning correlations between external inputs in my doctoral dissertation (1992) and related research on situation decomposition techniques when I met Professor Matsuo and we co-authored “Networks – Predictability – Attribute Generation” together. Through this connection, I have been a member of the Matsuo Lab since the summer of this year (2019).
    In the coming year, I hope to develop in these directions, contributing to the OS “World Models and Intelligence” @ JSAI 2020 and research strategy development, as well as academic activities related to general-purpose artificial intelligence and brain and AI.”

    2. Papers

    Matsuo Lab. started its research on web engineering with the aim of “creating “intelligence”” and has recently been working on deep generative modeling and deep reinforcement learning as the pillars of its research area. This year, our research has been adopted and received various awards at the following international conferences.

    ◆ICLR2019 Workshop (Limited Label Data), accepted to ECML PKDD


    Kei Akuzawa (1st year Ph.D.)
    *There will be no comments this year due to circumstances.

    WI2019 Best Student Paper Award

    Omi Nakagawa (1st year, Ph.D.)

    I am very honored to receive such a prestigious award.
    I am very honored to receive such a prestigious award, and I have had the opportunity to interact with people of various nationalities and research fields through the award and presentation.
    I will continue to work on my research and its social implementation so that I can achieve even greater results in the coming year.”

    ◆Selected to receive the Student Incentive Award for the 2019 National Conference on Artificial Intelligence

    Shohei Taniguchi (1st year master’s student)

    Since I was assigned to the Matsuo Lab last year as an undergraduate thesis student, I have been mainly engaged in basic research. When I was first assigned to the lab, I was almost an amateur in the field of deep learning and had to learn it from scratch. In the area of research, I was fortunate to receive the Student Incentive Award at the national conference of the Japanese Society for Artificial Intelligence for my presentation on my undergraduate thesis research.
    In the past year, many things have changed in the Matsuo Lab, especially in the area of basic research, with the arrival of Dr. Shane Gu of Google Brain as a visiting researcher and our full-fledged entry into robotics research. I am very proud to be a part of this project.
    Although deep learning is being applied in a variety of fields, there are still many unexplored areas, and I think it is a very interesting subject for research. I would like to continue my efforts in cooperation with other laboratory members to produce even greater research results in the future.”

    3. Collaborative Research

    Matsuo Lab is working hard on joint research with companies to advance the social implementation of Deep Learning and data science as much as possible. This year, we have started projects such as natural language processing in the legal field and image recognition in the nursing care field, to name but a few.

    ◆Mori Hamada & Matsumoto

    Notice of Joint Demonstration Research (Research on the Use of IT and AI in Legal Services) with Matsuo Laboratory of the University of Tokyo and Eliza Co.

    ◆I Study Co.

    Launched joint research with the University of Tokyo’s Matsuo Laboratory on the construction of an AI detection system for nursing care facilities.

    4. New appointment of Dr. Matsuo

    For Dr. Matsuo, the head of the Matsuo Laboratory, there have been changes in his position, including his appointment as a professor and as an outside director of SoftBank Corp.

    Appointed Professor, Graduate School of Engineering, The University of Tokyo

    Appointed as Director of SoftBank Group

    Yutaka Matsuo

    Press Release on New Director Nominees
    https://group.softbank/corp/news/press/sb/2019/20190521_01/

    Thank you very much for all the congratulatory messages and flowers!

    5. other

    DL4US Content Release

    Exercise content for DL4US, a course for training Deep Learning engineers, is now available free of charge.

    Created by the Matsuo Lab for engineers, it is implementation-oriented and practical, starting with image recognition and translation models and working up to generative models and reinforcement learning.

    For more information, click here.

    Establishment of a satellite laboratory in Mitutoyo City

    We set up a satellite laboratory in Mitutoyo City to search for ways to improve the work efficiency of local companies suffering from labor shortages.

    In the laboratory, technical college students participate in research on a project basis and propose solutions to the problems of participating companies.

    If you want to know more about Matsuo Lab’s activities Please also see the “Deep Learning Study Group Activity Report” if you want to know more about our activities.

    Matsuo-Iwasawa Laboratory, The University of Tokyo – Matsuo…

    DeepLearning Study Group hosted by Matsuo Lab] One year activity report meeting:https://weblab.t.u-tokyo.ac.jp/dlsummerevent_20190718/

    Hello, this is Shimizu from Public Relations. Recently, we held an activity report meeting to deepen exchanges mainly among members of the DL Reading Group and DLHacks, a study group led by members of the Matsuo Laboratory at the University of Tokyo who take turns giving presentations each week with the aim of deepening each other’s knowledge. (Introduction article…

    In addition, Matsuo Lab is currently looking for members to research and support together.

    ◆Community Manager

    Infrastructure Engineer

    Please contact us if you share our philosophy of “Creating Intelligence and Pioneering the Future” for other positions as well.

    All of us at Matsuo Lab wish you good health and continued growth for your company.
    Please have a happy New Year.
    We look forward to working with you in the New Year.