|
Abstract:
|
In this talk, we will describe a new probabilistic time series model called Hidden Process Models (HPMs), motivated by and applied to functional Magnetic Resonance Imaging (fMRI) data. HPMs model time series data using a set of latent processes, each of which has a spatial-temporal signature. Processes are triggered by known events with variable delays, and their signatures may overlap in space and/or time. HPMs provide a principled, probabilistic framework for simultaneously learning the contribution of each process to the observed data, as well as the timing and identities of each instantiated process. They also provide a framework for evaluating and selecting among competing models that assume different numbers and types of underlying processes. We will describe the HPM framework and its learning and inference algorithms, and present experimental results demonstrating its use.
|
