PLSO: A generative framework for decomposing nonstationary timeseries into piecewise stationary oscillatory components
share
To capture the slowly time-varying spectral content of real-world time series, a common paradigm is to partition the data into approximately stationary intervals and perform inference in the time-frequency domain. This approach, however, lacks a corresponding nonstationary time-domain generative model for the entire data and thus, time-domain inference, such as sampling from the posterior, occurs in each interval separately. This results in distortion/discontinuity around interval boundaries and, consequently, can lead to erroneous inferences based on any quantities derived from the posterior, such as the phase. To address these shortcomings, we propose the Piecewise Locally Stationary Oscillation (PLSO) generative model for decomposing time-series data with slowly time-varying spectra into several oscillatory, piecewise-stationary processes. PLSO, being a nonstationary time-domain generative model, enables inference on the entire time-series, without boundary effects, and, at the same time, provides a characterization of its time-varying spectral properties. We propose a novel two-stage inference algorithm that combines the classical Kalman filter and the recently-proposed accelerated proximal gradient algorithm to optimize the nonconvex Whittle likelihood from PLSO. We demonstrate these points through experiments on simulated data and real neural data from the rat and the human brain.
READ FULL TEXT
