Directed Graph Neural Networks for Ranking and Angular Synchronization

This talk will cover two applications of directed graph neural networks (GNNs) to signal recovery problems, ranking and angular synchronization. Recovering global rankings from pairwise comparisons has wide applications, such as sports team ranking. In the first half of the talk, I will introduce GNNRank, a trainable GNN-based framework with directed graph embedding, to tackle the ranking problem.  In the second half of the talk, I will apply directed GNNs to the angular synchronization problem, which aims to accurately estimate (up to a constant additive phase) a set of unknown angles from  noisy measurements of their offsets. Applications include, for example, sensor network localization, phase retrieval, and distributed clock synchronization. For both applications, new objectives will be discussed.

Bio
https://sherylhyx.github.io/