[3G3-OS-18a] 感情とAI(1)

[3G3-OS-18a-04] Extraction of Neuroscientific Findings by Visualization of Deep Neural Network

〇Kazuki Sakuma¹, Junya Morita¹, Taiki Nomura², Takatsugu Hirayama³, Yu Enokibori³, Kenji Mase³ (1. Faculty of Informatics, Shizuoka University, 2. Graduate School of Information Science, Nagoya University, 3. Graduate School of Informatics, Nagoya University)

Keywords:EEG, emotion, DNN

Recently, research using deep learning has been conducted in various fields. Additionally, research on visualization methods learned by deep learning has also been actively conducted. Furthermore, the relationship between the human subjective state and electroencephalogram (EEG) has been clarified in the psychophysiological field. In this research, we apply the visualization method developed in the image field to the analysis of EEG. Using this method, we examine whether we can abstract physiologically reasonable structure of brain activity from the network visualizing EEG signals.The result of our experiment indicated the two important brain structures showing consistency with the previous neuroscience studies. We consider that our proposed method has some utilities as a tool to progress scientific understanding of human mind.