Popis: |
The processing of signals on graph edges is challenging considering that Graph Signal Processing techniques are defined only on the graph nodes. Leveraging the Line Graph to transform a graph edge signal onto the node of its edge-to-vertex dual, we propose the Line Graph Least Mean Square (LGLMS) algorithm for online time-varying graph edge signal prediction. By setting up an $l_2$-norm optimization problem, LGLMS forms an adaptive algorithm as the graph edge analogy of the classical adaptive LMS algorithm. Additionally, the LGLMS inherits all the GSP concepts and techniques that can previously be deployed on the graph nodes, but without the need to redefine them on the graph edges. Experimenting with transportation graphs and meteorological graphs, with the signal observations having noisy and missing values, we confirmed that LGLMS is suitable for the online prediction of time-varying edge signals. |