Tomasz M. RUTKOWSKI, PhD
1. Cogent Labs Inc., Tokyo, Japan, https://www.cogent.co.jp/
2. Graduate School of Education, The University of Tokyo,
3. BCI-lab, Tokyo, Japan, http://tomek.bci-lab.info/
End-to-end Deep-learning Approaches for Online BCI and Offline Experiment Brainwave Analyses
The talk will first review commercial AI and deep learning models developed by Cogent Labs Inc. in Tokyo. The main part of the lecture will focus next on so-called end-to-end deep learning approaches for online and offline analysis of electrical signals (time-series) produced by the human brain (EEG) within brain computer interface (BCI) paradigms or educational psychology experiments. The end-to-end deep learning processing pipelines allow for fully data-driven and adaptive methods’ design (without expensive hand-crafted or engineered stages). The end-to-end deep learning methods are based on a training of models without imposing unnecessary assumptions or limitations, yet resulting in discovering of new meaningful features within multivariate brainwaves leading to more meaningful results. Real-time decoding of emotional or creative flow states, and education related cognitive states, which can be utilized to improve brain machine interactions and suggestions, are part of our research activities. In addition, while a traditional AI aspires to build autonomous intelligent computer systems, our BCI initiative is particularly interested in human intelligence augmentation (IA). The basic idea of IA is computers that do not replace human minds, but instead supplement, support and connect them with the AI-cloud. Current results of the end-to-end deep learning model application for BCI and offline EEG processing, as well as future research goals will summarize the talk.
|日時||2017/12/12(火) 15:00 - 16:00|