Mesochronal Structure Learning

Abstract

Standard time series structure learning algorithms assume that the measurement timescale is approximately the same as the timescale of the underlying (causal) system. In many scientific contexts, however, this assumption is violated: the measurement timescale can be substantially slower than the system timescale (i.e., many intermediate time series datapoints are missing). This assumption violation can lead to significant learning errors. In this paper, we provide a novel learning algorithm that can extract system-timescale structure given measurement data that undersample the underlying system. Substantial algorithmic optimizations were required to achieve computational tractability. We conclude by showing that the algorithm is highly reliable at extracting system-timescale structure from undersampled data.