Graph Neural Flows for Unveiling Systemic Interactions Among Irregularly Sampled Time Series
This addresses the challenge of accurately modeling systemic interactions for applications like time series classification and forecasting, though it appears incremental as it builds on existing graph and ODE techniques.
The paper tackles the problem of predicting dynamics in interacting systems with irregularly sampled time series by developing a graph-based model that learns conditional dependencies and integrates them with continuous-time ODEs, resulting in substantial enhancements over non-graph and other graph-based methods.
Interacting systems are prevalent in nature. It is challenging to accurately predict the dynamics of the system if its constituent components are analyzed independently. We develop a graph-based model that unveils the systemic interactions of time series observed at irregular time points, by using a directed acyclic graph to model the conditional dependencies (a form of causal notation) of the system components and learning this graph in tandem with a continuous-time model that parameterizes the solution curves of ordinary differential equations (ODEs). Our technique, a graph neural flow, leads to substantial enhancements over non-graph-based methods, as well as graph-based methods without the modeling of conditional dependencies. We validate our approach on several tasks, including time series classification and forecasting, to demonstrate its efficacy.