Robust Training of Temporal GNNs using Nearest Neighbours based Hard Negatives

arXiv:2402.09239v1 Announce Type: new
Abstract: Temporal graph neural networks Tgnn have exhibited state-of-art performance in future-link prediction tasks. Training of these TGNNs is enumerated by uniform random sampling based unsupervised loss. During training, in the context of a positive example, the loss is computed over uninformative negatives, which introduces redundancy and sub-optimal performance. In this paper, we propose modified unsupervised learning of Tgnn, by replacing the uniform negative sampling with importance-based negative sampling. We theoretically motivate and define the dynamically …

abstract art arxiv context cs.ir cs.lg example future gnns graph graph neural networks link prediction loss networks neural networks performance positive prediction random robust sampling state tasks temporal training type uniform unsupervised