We are working on incorporating white matter structural connectivity patterns of empirical brain networks into artificial neural network models and evaluating the performance of such models on various machine learning tasks. Through this research, we aim to elucidate the function of inter-areal brain network structure in solving real-world tasks and to develop artificial intelligence technologies that incorporate knowledge of empirical brain networks. [ongoing]

With discrete-event simulation techniques, we work on modeling how information processed in each cortical area of the brain propagates through the large-scale network of structural white matter connections. We are particularly interested in using the Internet metaphor for communication to investigate which general communication rule best approximates the dynamics of information propagation in structural brain networks. [ongoing]

Using graph-theoretic methods in network science, we work on characterizing dynamic fluctuations in global network properties of time-resolved functional brain connectivity. We also develop research to simulate these brain network fluctuations by constructing nonlinear dynamical systems and to reveal the generative mechanisms of these fluctuations by manipulating elements of the systems and studying their responses in simulations. [ongoing]

We combined statistical machine learning techniques (variational Bayes, sparse modeling, and state space modeling) to improve computational efficiency and estimation accuracy in solving high-dimensional dynamic inverse problems. Our approach allowed simultaneous reconstruction of the spatial distribution and temporal propagation of neural current sources across the whole brain from multimodal neuroimaging data (MEG, fMRI, and dMRI). [completed]