27-30 Jul., 2015, Osaka



MIRU2015

The 18th Meeting on Image Recognition and Understanding

Special Talk


Please note that all talks are presented in Japanese.


Special Talk 1

Yukiyasu Kamitani (Graduate School of Informatics, Kyoto University / ATR Computational Neuroscience Laboratories)
"Reading the mind by brain imaging"
Recent neuroimaging work is pushing in the direction of predictive science with the help of computational and machine learning modeling. Statistical pattern recognition algorithms have been applied to single-trial multivoxel patterns of fMRI data to make predictions about behavior and cognition including seen stimuli, motor intention, recalled memory, and dreamed contents, realizing a primitive form of “neural mind-reading”. The scientific approach using this method is now widely recognized as “multivoxl (multivariate) pattern analysis (MVPA)” or “brain decoding”. In this talk, I will present methodological principles and technical limitations of this approach, while highlighting the gap between the state of the art and what general people think “mind-reading” is like. I discuss new approaches that could help to fill the gap and enable us to read out a wider variety of mental states experienced in daily life.

<Profile>
Ph.D. Professor at Graduate School of Informatics, Kyoto University and Head of Department of Neuroinformatics at ATR Computational Neuroscience Laboratories, Kyoto, Japan. He received B.A. in Cognitive Science from University of Tokyo in 1993, M.S. in Philosophy of Science from University of Tokyo in 1995, and Ph.D. in Computation and Neural Systems from California Institute of Technology in 2001. He continued his research in cognitive and computational neuroscience at Harvard Medical School and Princeton University. In 2004, he joined ATR Computational Neuroscience Laboratories, where he heads Department of Neuroinformatics. In 2015, he started a new lab at Kyoto University. As a pioneer in brain decoding, he brought a new paradigm to brain imaging in which mental contents are predicted from brain activity patterns using machine learning models. He was named Research Leader in Neural Imaging on “Scientific American 50” in 2005. He has received several distinguished awards including Tsukahara Memorial Award (2013), and JSPS Award (2014).


Special Talk 2

Toshiki Kindo (Toyota Motor Corporation)
"Current status of autonomous driving technologies -TOYOTA's research activities-"
Autonomous driving is one of current hot topics in robotics and related areas. Autonomous driving development has be accelerated by DARPA grand challenge at 2004, which is a race of autonomous driving vehicles. Now not only automotive companies but also IT companies, Google etc., tackle the development of the technologies. In this talk, I introduce the impact of DARPA grand challenge and key technologies in autonomous driving based Toyota's research activities.

<Profile>
He obtained a B.A.(Science) in 1982 and an PhD(Physics) in 1987 from the Tohoku University. From 1985 to 1987 He studied at National Laboratory of High Energy Physics(KEK). He was a researcher at Japan Atomic Research Institute Tokai(1987-1989), Matsushita Research Institute Tokyo, Panasonic(1989-2003). He was double as a researcher of PRESTO of Japan Science and Technology Agency from 1995-2000. He joined Toyota Motor Corporation in 2003. His current research interests are in robotics including autonomous driving, artificial intelligence, neural networks, machine learning, information retrieval, and human-machine interaction to share knowledge and awareness.


Special Talk 3

Kentaro Inui (Tohoku University)
"Challenges for Modeling “Reading between the Lines”: Knowledge, Learning, Reasoning, and Grounding"
Natural-language processing (NLP) has witnessed rapid progress based on the success of statistical methods over the last 20 years. However, most present NLP technologies only deal with content explicitly mentioned in the sentences and context beyond the sentence level is not considered. Deep understanding of the contextual structure of text, i.e. “reading between the lines”, is still a long-term goal that requires many technological innovations. Fortunately, the "bottleneck of knowledge acquisition” is now becoming resolvable through the automatic acquisition of linguistic and world knowledge from huge text data. If large-scale knowledge bases including a broad range of commonsense become available in the near future, what will be the remaining issues? In this talk, I will provide an overview of the current frontier of NLP, featuring knowledge acquisition, representation learning, commonsense reasoning, and the integration of language understanding and computer vision, and I will argue their future prospects.

<Profile>
Kentaro Inui. He is a Professor at the Graduate School of Information Sciences at Tohoku University, where he leads the Communication Science Lab. He received his doctorate degree in Engineering from Tokyo Institute of Technology in 1995. After that he worked as a Research Associate at Tokyo Institute of Technology and worked as an Associate Professor at Kyushu Institute of Technology. In 2002 he moved to Nara Institute of Science and Technology, working as an Associate Professor until he joined Tohoku University in 2010. His general areas of research are natural language processing and artificial intelligence. He currently serves as the Chair of IPSJ-SIG Natural Language Processing.