International Exchange of Ideas at Boundary of Computer Science and Neuroscience
In a verdant setting in the University of Tokyo Hongo campus, overlooking a forested pond and with the meditative intensity of a neighboring Zen archery studio, scientists from 16 countries gathered to learn from global experts how to create synergy between two of the hottest research fields in the last decade, computer science and neuroscience.

With 29 international students, 69 UTokyo students and researchers, and 18 teaching faculty, the inaugural IRCN Neuro-inspired Computation Course transpired in 4 days from March 21-24, 2019 at Sanjo Kaikan Hall and nearby IRCN meeting spaces. The course consisted of lectures interleaved with posters and breakout discussion sessions.

The students heard from a broad expanse of frontier fields including brain and computer architecture, dynamical neural networks, machine and deep learning, brain development and disorders, and reinforcement learning. Many lecturers covered multiple areas giving an interdisciplinary focus to the proceedings and enabling bridging between disciplines.

“There are people who work on everything here. It's not only that you have speakers in different fields but they work across different fields. They are used to switching”.
Pau Vilimelis Aceituno, PhD student, Max Planck Institute for Mathematics in the Sciences


The course was the brainchild of IRCN Director Takao Hensch, who has long pursued his landmark research on the developmental physiology of visual cortex in collaboration with computational neuroscientists. Hensch attributes his approach to science and computing to his mentor at The University of Tokyo and RIKEN, the late Professor Masao Ito.

Day 1 included student introductions and lectures in brain architecture and brain dynamics. Partha Mitra delivered a riveting opening talk that captured the current excitement surrounding the fusion of neuroscience and AI fields, and his personal journey from physics to mapping brain networks with high-resolution microscopy.
“The difference [between artificial and real neurons is that the real] neurons have spontaneous dynamics. If they are sitting there, they are doing something.”
Partha Mitra, Cold Spring Harbor Laboratory


Markus Diesmann talked about remarkable recent progress in supercomputing of brain simulations and the role of neuromorphic computing in building realistic architectures. Stefano Panzeri devoted his lecture to modeling of neural networks based on data and theory while Kazuyuki Aihara shared his passion for mathematical approaches to brain.

“Our work is about the mathematical modeling and the analysis of dynamics in neural networks, and how to apply these models for prediction in healthy and disease brains.”
Kazuyuki Aihara, The University of Tokyo


Day 2 began with an introduction to the role of computation in brain development and disorders by Taro Toyoizumi and Arvind Kumar. The highly anticipated afternoon session was led by Surya Ganguli on the new generation of human-directed AI and the evolution of computer-driven robotics applications by Yasuo Kuniyoshi and Jun Tani.

Learning models were the topic of Day 3 Kenji Doya and Daniel Brunner described the current interest in reinforcement learning, while Graham Taylor and Masashi Sugiyama covered the prospects in deep machine learning for building better performing machines.

Three poster sessions allowed time for interaction between course participants, and on Day 1 and Day 3 students met with breakout session mentors Jon Schneider and Michele McCarthy, and heard a lecture by Nima Dehgani, to gain experience on Day 4 with building models and envisioning team collaboration by assembling reports on potential team projects.

“I got to talk to so many people who, like me, were trying to move from one field to another. Mapping between the machine learning and brain science area, I would like to try something like that in my institute.”
Sandyha Tripathi, PhD student, Indian Institute of Technology-Mumbai


There was poster prize sponsored by the journal Frontiers in Neural Circuits. The winners were Andrea Navas-Olive of the Cajal Institute, Luziwei Leng from the University of Heidelberg, and Yiqiao Wang of the Karolinska Institute. Every poster contributed greatly to the intellectual diversity of the course and cross-field learning.

Course participants also had time for fun, with many visiting Japan for the first time. Trips around Tokyo such as late night “electric town” Akihabara, the 5 am Tsukiji fish market tour, IRCN cruise on the Sumida river past SkyTree tower, and Imperial Palace. Warm weather and early cherry blossoms lent a seasonal coloring to the Tokyo backdrop.

The students and lecturers agreed that the course was an important step toward raising awareness for neuro-inspired AI, leveraging the remarkable efficiency of the human brain that current AI cannot touch. Building from principles of brain development, IRCN will help researchers around the world work together to build novel AI for science and society.
Please check the IRCN website for the next course announcement. Email: course@ircn.jp
For more coverage: Visit Twitter @IRCN_UTokyo

Writing: Charles Yokoyama
Reporting: Sara El-Shawa
Tweeting: Walid Yassin

Course Topics and Lecturers

Brain Architecture
Partha Mitra (Cold Spring Harbor Laboratory, USA) - Abstract&Bio
Markus Diesmann (Jülich Research Center, Germany) - Abstract&Bio
Brain Dynamics
Kazuyuki Aihara (IRCN/The University of Tokyo, Japan) - Abstract&Bio
Stefano Panzeri (Italian Institutes of Technology, Italy) - Abstract&Bio
Michelle McCarthy (Boston University, USA)
Machine Learning
Masashi Sugiyama (RIKEN and IRCN/The University of Tokyo, Japan) - Abstract&Bio
Graham Taylor (University of Guelph and Vector Institute/Google Brain Montreal, Canada) - Abstract&Bio
Jonathan Schneider (University of Toronto, Canada)
Dynamical Systems
Surya Ganguli (Stanford University, USA) - Abstract&Bio
Jun Tani (Okinawa Institute of Science and Technology, Japan) - Abstract&Bio
Yasuo Kuniyoshi (IRCN/The University of Tokyo, Japan) - Abstract&Bio
Reinforcement Learning
Kenji Doya (Okinawa Institute of Science and Technology, Japan) - Abstract&Bio
Daniel Brunner (CNRS, France) - Abstract&Bio
Brain Development / Disorders
Taro Toyoizumi (RIKEN Center for Brain Science, Japan) - Abstract&Bio
Arvind Kumar (KTH Royal Institute of Technology, Sweden) - Abstract&Bio
Nima Dehghani (Massachusetts Institute of Technology, USA)

Course Details

Date
March 21 to 24, 2019
Venue
Sanjo-Kaikan Hall, The University of Tokyo,
Hongo, Bunkyo-ku, Tokyo 113-0033, JAPAN
Detailed Program
Refer to the PDF




Co-Supported by Next Generation Artificial Intelligence Research Center, KAKENHI Project on Artificial Intelligence and Brain Science and Japan Agency for Medical Research and Development