|
Abstract:
|
Neuroscience, social network analysis and many other disciplines involve the generation of multivariate count data observed over time. For example, spike trains at multiple locations, or events such as liking Facebook posts correspond to multivariate temporal data where typically the number of time series is extremely large. One of the important questions in this setting is how different time series (i.e. voxels in the brain or people in a social network) influence each other. In this talk, I present recent results on how this challenge can be addressed using high-dimensional generalized linear auto-regressive and moving average models. I provide both implementable algorithms and theoretical guarantees which support our generalized linear auto-regressive model and its modifications to suitable applications.
|
