Temporal Generalization Estimation in Evolving Graphs

arXiv:2404.04969v1 Announce Type: new
Abstract: Graph Neural Networks (GNNs) are widely deployed in vast fields, but they often struggle to maintain accurate representations as graphs evolve. We theoretically establish a lower bound, proving that under mild conditions, representation distortion inevitably occurs over time. To estimate the temporal distortion without human annotation after deployment, one naive approach is to pre-train a recurrent model (e.g., RNN) before deployment and use this model afterwards, but the estimation is far from satisfactory. In this …

abstract annotation arxiv cs.ai cs.lg deployment fields gnns graph graph neural networks graphs human networks neural networks representation struggle temporal type vast