Robust Training of Temporal GNNs using Nearest Neighbours based Hard Negatives
Not on the map. Why?
Abstract (arXiv)
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 computed distribution for a sampling of negative examples. Finally, using empirical evaluations over three real-world datasets, we show that Tgnn trained using loss based on proposed negative sampling provides consistent superior performance.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 35 of its 57 references that are in the map. First posted on arXiv 14 Feb 2024.
Builds on
35 of 57 references.
References · 35
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
- 2017Graph Attention Networkscited by 8,854
- 2017Inductive Representation Learning on Large Graphscited by 6,284
- 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,779
- 2014On the Properties of Neural Machine Translation: Encoder-Decoder Approachescited by 1,303
- 2018Graph Convolutional Neural Networks for Web-Scale Recommender Systemscited by 1,133
- 2018Deep Graph Infomaxcited by 1,101
- 2020Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrievalcited by 846
- 2015Image-based Recommendations on Styles and Substitutescited by 702
- 2020Contrastive Learning with Hard Negative Samplescited by 619
- 2020A Survey on Contrastive Self-supervised Learningcited by 553
- 2020Deep Graph Contrastive Representation Learningcited by 535
23 more
- 2020Inductive Representation Learning on Temporal Graphscited by 454
- 2019EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphscited by 443
- 2019Predicting Dynamic Embedding Trajectory in Temporal Interaction Networkscited by 443
- 2020RocketQA: An Optimized Training Approach to Dense Passage Retrieval for Open-Domain Question Answeringcited by 410
- 2020Hard Negative Mixing for Contrastive Learningcited by 399
- 2018Not All Samples Are Created Equal: Deep Learning with Importance Samplingcited by 323
- 2021Efficiently Teaching an Effective Dense Retriever with Balanced Topic Aware Samplingcited by 238
- 2020Graph Representation Learning via Graphical Mutual Information Maximizationcited by 217
- 2019Time2Vec: Learning a Vector Representation of Timecited by 208
- 2021RocketQAv2: A Joint Training Method for Dense Passage Retrieval and Passage Re-rankingcited by 175
- 2021Optimizing Dense Retrieval Model Training with Hard Negativescited by 154
- 2021Self-supervised Graph-level Representation Learning with Local and Global Structurecited by 125
- 2018DynGEM: Deep Embedding Method for Dynamic Graphscited by 100
- 2020Contrastive Learning with Adversarial Examplescited by 91
- 2018Community Interaction and Conflict on the Webcited by 72
- 2020Understanding Negative Sampling in Graph Representation Learningcited by 66
- 2019Temporal Network Embedding with Micro- and Macro-dynamicscited by 51
- 2019Node Embedding over Temporal Graphscited by 45
- 2022TIGGER: Scalable Generative Modelling for Temporal Interaction Graphscited by 18
- 2023GRAFENNE: Learning on Graphs with Heterogeneous and Dynamic Feature Setscited by 7
- 2020Privacy-Aware Recommender Systems Challenge on Twitter's Home Timelinecited by 4
- 2022A Survey on Temporal Graph Representation Learning and Generative Modelingcited by 4
- 2021Structure-Aware Hard Negative Mining for Heterogeneous Graph Contrastive Learningcited by 3
Built on by
Citing papers per year: 2025: 1.
1 paper.