arXiv 1905.09944
Unsupervised Discovery of Temporal Structure in Noisy Data with Dynamical Components Analysis
By David G. Clark, Jesse A. Livezey, et al.
Published 2019-05-23
Wiki summary
Explore the paper's summary, context, and related research on Papiers.
Linear dimensionality reduction methods are commonly used to extract low-dimensional structure from high-dimensional data. However, popular methods disregard temporal structure, rendering them prone to extracting noise rather than meaningful dynamics when applied to time series data. At the same time, many successful unsupervised learning methods for temporal, sequential and spatial data extract features which are p…