Search papers, labs, and topics across Lattice.
The paper introduces ODEBRAIN, a Neural ODE framework for modeling continuous-time EEG dynamics by integrating spatio-temporal-frequency features into spectral graph nodes. This approach addresses limitations of recurrent architectures in capturing instantaneous, nonlinear EEG characteristics and mitigating compounded prediction errors. Experiments demonstrate that ODEBRAIN significantly improves EEG dynamics forecasting with enhanced robustness and generalization compared to existing methods.
By modeling EEG dynamics in continuous time with Neural ODEs, ODEBrain avoids error accumulation and captures instantaneous, nonlinear brain activity better than recurrent architectures.
Modeling neural population dynamics is crucial for foundational neuroscientific research and various clinical applications. Conventional latent variable methods typically model continuous brain dynamics through discretizing time with recurrent architecture, which necessarily results in compounded cumulative prediction errors and failure of capturing instantaneous, nonlinear characteristics of EEGs. We propose ODEBRAIN, a Neural ODE latent dynamic forecasting framework to overcome these challenges by integrating spatio-temporal-frequency features into spectral graph nodes, followed by a Neural ODE modeling the continuous latent dynamics. Our design ensures that latent representations can capture stochastic variations of complex brain states at any given time point. Extensive experiments verify that ODEBRAIN can improve significantly over existing methods in forecasting EEG dynamics with enhanced robustness and generalization capabilities.
