Research attempt of computational graph-based neurolinguistics integrating brain fMRI, machine learning and complex networks

2014 Fiscal Year Final Research Report

• Project/Area Number: 23500171
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent informatics
• Research Institution: Tokyo Institute of Technology
• Principal Investigator: AKAMA Hiroyuki 東京工業大学, 社会理工学研究科, 准教授 (60242301)
• Co-Investigator(Kenkyū-buntansha): TOKOSUMI Akifumi 東京工業大学, 大学院社会理工学研究科, 教授 (50125332)
• Project Period (FY): 2011-04-28 – 2015-03-31
• Keywords: 脳科学 / fMRI / 機械学習 / 複雑ネットワーク / 自然言語処理

Outline of Final Research Achievements

We carried out an brain informatics study by performing fMRI experiments, conducting a statistical analysis of the datasets gathered during a covert word property generation task, especially using our original machine learning to classify and predict various cognitive states. Addressing the problem of vulnerability in Multi-Voxel Pattern Analysis (MVPA) eliciting a considerable performance penalty (decrease in accuracy) when executed across sessions or participants, we executed a depthful fitting analysis focusing on cross-modal and individual variation that most previous research has not involved so far. As a result, a key to success in cross-learning on semantic cognition was revealed by underscoring the significance of feature selection strategy founded upon integrated spatio-temporal patterns of neural activity. Furthermore, we conducted experiments involving language switching, with early bilingual speakers and extracted some informative regions using both GLM methods and MVPA.

Free Research Field

知能情報学