Abstract:Dynamic systems of graph signals are encountered in various applications, including social networks, power grids, and transportation. While such systems can often be described as state space (SS) models, tracking graph signals via conventional tools based on the Kalman filter (KF) and its variants is typically challenging. This is due to the nonlinearity, high dimensionality, irregularity of the domain, and complex modeling associated with real-world dynamic systems of graph signals. In this work, we study the tracking of graph signals using a hybrid model-based/data-driven approach. We develop the GSP-KalmanNet, which tracks the hidden graphical states from the graphical measurements by jointly leveraging graph signal processing (GSP) tools and deep learning (DL) techniques. The derivations of the GSP-KalmanNet are based on extending the KF to exploit the inherent graph structure via designing a graph frequency domain filtering and replacing the Kalman gain (KG) with a graph filter that minimizes the prediction error. Thus, it considerably simplifies the computational complexity entailed in processing high-dimensional signals and increases the robustness to small topology changes. Then, we use data to learn the KG, namely, the graph filter, following the recently proposed KalmanNet framework, which copes with partial and approximated modeling, without forcing a specific model over the noise statistics. Restricting the KG to a graph filter in the proposed GSP-KalmanNet reduces learned parameters, thereby enhancing stability. Our empirical results demonstrate that the GSP-KalmanNet achieves enhanced accuracy and run time performance, and improved robustness to model misspecifications compared with both model-based and data-driven benchmarks.

Sampling Policy Design for Tracking Time-Varying Graph Signals with Adaptive Budget Allocation

Dynamic Sampling Allocation and Design Selection.

On Critical Sampling of Time-Vertex Graph Signals

Online Signed Sampling of Bandlimited Graph Signals

Stochastic Control Framework for Determining Feasible Alternatives in Sampling Allocation

Sequential Sampling and Estimation of Approximately Bandlimited Graph Signals

Adaptive Graph Signal Processing: Algorithms and Optimal Sampling Strategies

Fast Graph Sampling Set Selection Using Gershgorin Disc Alignment

The Smart Sampling Kalman Filter with Symmetric Samples

Subset Random Sampling of Finite Time-vertex Graph Signals

The Design of Asynchronous Kalman Filter-based Tracking Loop in Digital Domain

Bayesian Design of Sampling Set for Bandlimited Graph Signals

Graph-Based Signal Sampling with Adaptive Subspace Reconstruction for Spatially-Irregular Sensor Data

Target Tracking Algorithm Based on Adaptive Strong Tracking Extended Kalman Filter

Dynamic Sampling For Feasibility Determination

Efficient Directed Graph Sampling via Gershgorin Disc Alignment

GSP-KalmanNet: Tracking Graph Signals via Neural-Aided Kalman Filtering

Active Sampling For Approximately Bandlimited Graph Signals

Practical graph signal sampling with log-linear size scaling

Real-time Kinematic Positioning Algorithm in Graphical State Space