JNNS President’s Greeting
I am very honored to serve as president of the Japanese Neural Network Society. In accepting this role, I was blessed to think back on the steps I’ve taken in my own research and to ponder what and how to contribute to the research community and to society as a whole. The year 1989, which is when the Japanese Neural Network Society was founded, was also the year I started my research on computational neuroscience. At that time, since I had studied robotics at university, I was very interested in artificial intelligence and theoretical frameworks of adaptive control. As such, I was enthusiastic about studying neural networks, which share many similarities with adaptive control. Since then, the boom of neural networks, which had once faded down, has now been gaining momentum again in recent years. Our society journal, “Neural Networks” , is currently one of the most high profile journals in the field. The reason for this boom is of course not only the fundamental attractiveness of brain computation but also the continuous theoretical examinations of machine learning in the 90s and subsequent dramatically improved computer performance. Thanks to these trends and to the massive data that is now accessible due to the worldwide Internet, many cutting-edge applications such as automatic visual feature extraction and champion Igo-AI are being trained by reinforcement learning. The core technology underlying AI today is really neural network learning. Many of you may know that stochastic gradient descent, which was developed by Dr. Shunichi Amari, and the neocognitron, which was developed by Dr. Kunihiko Fukushima, both of whom are founders of this society, are the roots of deep neural network learning technology . What the history of the neural network research teaches us is the importance of pondering the central questions in our field and in society and of continuing to develop what we are really interested in.
The Japanese Neural Network Society focuses not just on models of neural circuits and machine learning in a narrow sense but rather is an interdisciplinary community where a wide range of researchers gather to understand the principles and mechanisms of complex brain information processing for sensation, perception, motor control, emotion, consciousness, etc. and to devise general schemes and build frameworks for various applications by using some analogy from neural mechanisms. Many of our members also have expertise in related fields such as neuroscience, physiology, psychology, biology, mathematics, informatics, electrical and electronic engineering, mechanical engineering, and physics. We would like to help and encourage all society members to engage in discussions and communications with other members in heterogeneous fields.
Last year was a very difficult one for universities and research institutes due to the unprecedented COVID-19 disaster. Most of us were not even allowed to go to our own laboratories, experiments could not be conducted as before, and all classes and research meetings had to be moved online. In addition, many conferences were cancelled or held online. In fact, last year’s JNNS annual conference could not be held as planned, and was forced to move online. While the rapid development of net-conferencing systems has been quite helpful, many people might not be familiar or adept with such styles of communication. On the other hand, there are also some advantages to this situation, such as saving time and effort with regard to traveling, being able to participate casually, and being able to listen to talks many times if they are recorded. Actually, I had an excellent experience in attending the international symposium of Artificial Intelligence and Brain Science , which was held online. Before the online discussion session, it was quite helpful for me to re-watch several videos of the talks, as they were slightly different from my current field. Recently, there have been signs of improvement in the pandemic situation thanks to widespread vaccination, but it seems we will have to continue to be careful about preventing infection for some time to come. To avoid risks, we plan to hold the annual conference online this year, too. The conference executive committee will make plans so that everyone can enjoy the presentations, talks, and discussions. In addition, the society committee will endeavor to introduce new ideas and methods to make the society more attractive to students and researchers. Together, let’s refresh our minds and move forward so that we can turn crisis into opportunity and develop our research with excitement.
・Ko Sakai (President 2015-2016)
What comes to mind at the word neural networks? One may think of a variety of things from this word. Progress of neuroscience in the past decades has revealed what are neurons and how they connect to others. By combining such knowledge, we can analyze and predict networks of neurons in the brain. We can also investigate what network is necessary to realize a particular brain function such as perception and cognition. Even our mind and intention may be generated by neural networks. These are perhaps the most intrinsic parts of the study of neural networks.
Neural networks may not be limited to neuroscience, physiology and anatomy. Neural networks include models of neural systems with the aim of understanding principles of neural functions and mechanisms including cognition, memory, intention, coding scheme, information processing and others. In parallel with understanding neural systems, the study of neural networks would propose innovative and advantageous principles and methods in mathematical sciences and engineering. Excellent examples may include neural networks in nineties and a recent development in deep learning.
The study of neural networks is frontier and interdisciplinary which covers a wide range of fields from physiology and psychology to information science, mathematical science, statistics and engineering. Our Neural Network Society was established in 1989 to promote the study of neural networks with the specific aims of introducing advances of research and exchanging and spreading knowledge. The society organizes conferences, workshops and tutorials, publishes magazines, promote new fields, and supports young scientists and engineers. We also contribute to international communities by editing and publishing the Journal, Neural Networks, in cooperation with International Neural Network Society and European Neural Network Society, and co-organizing International Conference On Neural Information Processing as a member of Asia Pacific Neural Network Assembly.
If you find a interest in neural networks, please join one of our activities. We are looking forward to seeing you and talking about neural networks.