Please register for the CV & Online Identity Workshop below.

I will present the "energy landscape analysis" of multivariate time series data. In this analysis, one identifies the state of the system at each time point as the position of a "ball" constrained on an energy landscape inferred from data. The energy landscape is constructed using the inverse Ising model (also called Boltzmann machine) and can be regarded as a representation of a correlational network inferred from the data. A ball tends to go downhill on the energy landscape whereas it sometimes goes uphill to transit from one local minimum of the energy to another, possibly corresponding to major dynamical transitions of the system. The application of the method to neuroimaging data is illustrated (while the method is not domain specific). I will also present a further development of the method with which to estimate how close the time series to the criticality and its application to the "brain criticality hypothesis".