pith. sign in

arxiv: 2006.10637 · v3 · pith:EERPHK5Unew · submitted 2020-06-18 · 💻 cs.LG · stat.ML

Temporal Graph Networks for Deep Learning on Dynamic Graphs

Emanuele Rossi , Ben Chamberlain , Fabrizio Frasca , Davide Eynard , Federico Monti , Michael Bronstein This is my paper

Pith reviewed 2026-05-17 16:58 UTC · model grok-4.3

classification 💻 cs.LG stat.ML
keywords temporal graph networksdynamic graphsgraph neural networksmemory moduleslink predictiontemporal dependenciesinductive learning
0
0 comments X

The pith

Temporal Graph Networks combine memory modules with graph operators to outperform prior dynamic graph methods while using less computation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper introduces Temporal Graph Networks as a general framework for deep learning on graphs that evolve over time through changing connections or node features. It models such graphs as ordered sequences of timed events and equips each node with a memory that records past interactions, which graph operators then combine with current structure to produce updated embeddings. The resulting model unifies several earlier dynamic graph techniques as special cases of its components and achieves higher accuracy on link prediction and related tasks. A sympathetic reader would care because many practical systems, from social interactions to biological processes, involve time-varying relations that static graph models cannot capture. The authors also show through experiments that the approach runs faster than previous specialized methods.

Core claim

Temporal Graph Networks are a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. Several previous models for learning on dynamic graphs can be cast as specific instances of this framework, and a detailed ablation identifies the configuration that reaches state-of-the-art results on transductive and inductive prediction tasks.

What carries the argument

Temporal Graph Network architecture, which maintains a memory state for each node that is updated via messages from timed events and then applies graph neural network operators on the current graph to compute embeddings for prediction.

If this is right

  • Several existing dynamic graph models can be expressed as specific choices of memory update and graph operator functions inside the TGN framework.
  • The optimal TGN configuration achieves state-of-the-art results on both transductive and inductive tasks for dynamic graph prediction.
  • TGNs deliver higher accuracy than prior methods at lower computational cost across the evaluated benchmarks.
  • The framework supports both transductive and inductive learning settings on evolving graphs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because TGNs subsume prior models, targeted improvements to the memory update rule could simultaneously advance performance across many existing dynamic graph techniques.
  • The efficiency gains could enable scaling temporal graph learning to larger event streams in domains such as recommendation systems or particle physics simulations referenced in the paper.
  • Exploring alternative memory architectures or operator choices beyond the ablated configurations might reveal further accuracy or speed trade-offs on new datasets.

Load-bearing premise

Dynamic graphs can be represented effectively as sequences of timed events such that memory modules and graph operators will capture the necessary temporal dependencies without prohibitive computational or modeling limitations.

What would settle it

Running the authors' best TGN configuration on the paper's benchmark datasets and observing that it fails to exceed the accuracy or speed of the strongest prior method on at least one transductive or inductive task would falsify the performance and efficiency claims.

read the original abstract

Graph Neural Networks (GNNs) have recently become increasingly popular due to their ability to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems. Despite the plethora of different models for deep learning on graphs, few approaches have been proposed thus far for dealing with graphs that present some sort of dynamic nature (e.g. evolving features or connectivity over time). In this paper, we present Temporal Graph Networks (TGNs), a generic, efficient framework for deep learning on dynamic graphs represented as sequences of timed events. Thanks to a novel combination of memory modules and graph-based operators, TGNs are able to significantly outperform previous approaches being at the same time more computationally efficient. We furthermore show that several previous models for learning on dynamic graphs can be cast as specific instances of our framework. We perform a detailed ablation study of different components of our framework and devise the best configuration that achieves state-of-the-art performance on several transductive and inductive prediction tasks for dynamic graphs.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces Temporal Graph Networks (TGNs), a generic framework for deep learning on dynamic graphs represented as sequences of timed events. It combines per-node memory modules updated via message passing with graph-based operators (e.g., attention or mean aggregation) and time encodings to capture temporal dependencies. The authors claim that this combination yields significant outperformance over prior methods while improving computational efficiency, demonstrate that several existing dynamic graph models are special cases of the TGN framework, and report an ablation study identifying a configuration that achieves state-of-the-art results on transductive and inductive link prediction tasks.

Significance. If the empirical claims hold, the work supplies a unifying and modular framework that clarifies relationships among prior dynamic graph methods and offers practical efficiency gains for applications in social networks, recommendation systems, and biology. The explicit casting of previous models as TGN instances and the component-wise ablation are constructive contributions that aid reproducibility and future extensions.

major comments (2)
  1. [§3] §3 (Framework description): The central efficiency and outperformance claims rest on fixed-size memory modules plus time encodings retaining relevant history across arbitrary-length event sequences while keeping per-event cost sub-quadratic. No analysis of information loss, compression artifacts, or scaling behavior under high event density or long horizons is provided, which directly bears on whether the memory update mechanism supports the stated advantages.
  2. [§5.3] §5.3 (Ablation study): The ablation varies memory, operator, and encoding choices on standard benchmarks yet omits controlled experiments on regimes with high event density or extended temporal horizons. Without these, it is impossible to verify that the reported gains and efficiency improvements generalize beyond the tested settings or that sequential memory updates avoid hidden costs relative to baselines.
minor comments (2)
  1. [Abstract] Abstract: The sentence 'significantly outperform previous approaches being at the same time more computationally efficient' is grammatically awkward; rephrase for clarity.
  2. [§4] §4 (Implementation details): Adding a table that reports wall-clock time and memory usage per event across varying sequence lengths would strengthen the efficiency claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major comment below and indicate the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: [§3] §3 (Framework description): The central efficiency and outperformance claims rest on fixed-size memory modules plus time encodings retaining relevant history across arbitrary-length event sequences while keeping per-event cost sub-quadratic. No analysis of information loss, compression artifacts, or scaling behavior under high event density or long horizons is provided, which directly bears on whether the memory update mechanism supports the stated advantages.

    Authors: We agree that a dedicated analysis of information retention and potential compression effects in the fixed-size memory would strengthen the theoretical justification for the efficiency claims. The current manuscript emphasizes empirical performance on real-world sequences of varying length, but does not include a formal treatment of long-horizon scaling or information loss. In the revised version we will add a short discussion subsection in §3 that derives the per-event memory update cost and comments on the conditions under which the fixed-size state is expected to preserve relevant history, referencing the message-passing and time-encoding design choices. revision: yes

  2. Referee: [§5.3] §5.3 (Ablation study): The ablation varies memory, operator, and encoding choices on standard benchmarks yet omits controlled experiments on regimes with high event density or extended temporal horizons. Without these, it is impossible to verify that the reported gains and efficiency improvements generalize beyond the tested settings or that sequential memory updates avoid hidden costs relative to baselines.

    Authors: We acknowledge that the ablation study is performed on the standard benchmark datasets and does not contain separate controlled experiments that systematically vary event density or horizon length. While the chosen datasets already exhibit a range of temporal densities and sequence lengths, we agree that targeted synthetic regimes would make the generalization argument more robust. In the revised manuscript we will add a short appendix with controlled experiments on synthetic graphs that increase event density and extend the temporal horizon, reporting both accuracy and wall-clock time relative to the same baselines used in the main ablation. revision: yes

Circularity Check

0 steps flagged

No circularity: TGN framework introduced as novel combination with empirical validation

full rationale

The paper presents TGNs as a new generic framework for dynamic graphs modeled as timed event sequences, combining memory modules with graph operators like attention or aggregation. Performance and efficiency claims rest on experimental results and ablations on standard benchmarks rather than any closed-form derivation or prediction that reduces to author-defined inputs by construction. Previous models are shown as special cases of the framework, which constitutes unification rather than renaming or self-referential fitting. No equations or steps in the abstract or description exhibit self-definitional, fitted-input, or self-citation load-bearing circularity; the central results are externally falsifiable via benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that dynamic graphs are well-modeled by timed event sequences and that memory modules can store sufficient history; no free parameters or invented entities are explicitly quantified in the abstract.

axioms (1)
  • domain assumption Dynamic graphs can be represented as sequences of timed events
    Explicitly stated in the abstract as the input representation for the framework.
invented entities (1)
  • Memory modules no independent evidence
    purpose: Store historical node information to enable temporal reasoning in dynamic graphs
    Introduced as a core component of TGNs without external validation or falsifiable prediction provided in the abstract.

pith-pipeline@v0.9.0 · 5488 in / 1258 out tokens · 68182 ms · 2026-05-17T16:58:39.051759+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 23 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Decoupled and Divergence-Conditioned Prompt for Multi-domain Dynamic Graph Foundation Models

    cs.LG 2026-05 conditional novelty 7.0

    DyGFM introduces decoupled pre-training and divergence-conditioned prompts to create the first multi-domain dynamic graph foundation model that outperforms baselines on node classification and link prediction.

  2. DRIFT: A Benchmark for Task-Free Continual Graph Learning with Continuous Distribution Shifts

    cs.LG 2026-05 accept novelty 7.0

    DRIFT is a benchmark for task-free continual graph learning under continuous distribution shifts, demonstrating that standard methods degrade without task boundary information.

  3. ATLAS: Efficient Out-of-Core Inference for Billion-Scale Graph Neural Networks

    cs.DC 2026-05 unverdicted novelty 7.0

    ATLAS achieves 12-30x faster out-of-core full-graph GNN inference on graphs up to 4B edges by switching to broadcast-based layer-wise execution with graph reordering, minimum-pending-message eviction, and GPU-accelera...

  4. Explaining Temporal Graph Predictions With Shapley Values

    cs.LG 2026-04 unverdicted novelty 7.0

    Event-level Shapley and feature-level Owen-value explainers for TGNNs outperform prior methods on metrics and datasets while revealing a timestamp extraction bug in TGAT.

  5. TRAVELFRAUDBENCH: A Configurable Evaluation Framework for GNN Fraud Ring Detection in Travel Networks

    cs.LG 2026-04 unverdicted novelty 7.0

    TravelFraudBench is a new configurable benchmark for GNN-based fraud ring detection in travel networks, simulating star, clique, and chain topologies and showing GraphSAGE outperforming MLP baselines on AUC and ring recovery.

  6. Beyond Nodes vs. Edges: A Multi-View Fusion Framework for Provenance-Based Intrusion Detection

    cs.CR 2026-04 unverdicted novelty 7.0

    PROVFUSION fuses three complementary views of provenance data with lightweight schemes and voting to achieve higher detection accuracy and lower false positives than node- or edge-only baselines on nine benchmarks.

  7. ChronoSpike: An Adaptive Spiking Graph Neural Network for Dynamic Graphs

    cs.LG 2026-02 unverdicted novelty 7.0

    ChronoSpike is a spiking GNN that integrates adaptive LIF neurons with spatial attention and temporal transformers to outperform baselines on dynamic graph benchmarks by 2% F1 while training 3-10x faster with fixed pa...

  8. Unsupervised Learning of Local Updates for Maximum Independent Set in Dynamic Graphs

    cs.LG 2025-05 unverdicted novelty 7.0

    Unsupervised GNN model learns local updates for approximate MaxIS on dynamic graphs, achieving competitive ratios on 200-1000 node instances and 1.00-1.18x larger solutions than other unsupervised models when generali...

  9. Graph Retention Networks for Dynamic Graphs

    cs.LG 2024-11 unverdicted novelty 7.0

    Graph Retention Networks extend retention to dynamic graphs to enable parallelizable training, O(1) inference, and chunkwise long-term training while delivering competitive performance with major efficiency gains.

  10. Attention Dispersion in Dynamic Graph Transformers: Diagnosis and a Transferable Fix

    cs.LG 2026-05 unverdicted novelty 6.0

    Diagnoses attention dispersion in CTDG Transformers under temporal shift and introduces differential attention to suppress common signals and achieve SOTA on shifted benchmarks.

  11. DRIFT: A Benchmark for Task-Free Continual Graph Learning with Continuous Distribution Shifts

    cs.LG 2026-05 unverdicted novelty 6.0

    DRIFT benchmark shows substantial performance degradation for continual graph learning methods under task-free continuous distribution shifts modeled via Gaussian mixtures.

  12. Predicting Channel Closures in the Lightning Network with Machine Learning

    cs.LG 2026-05 unverdicted novelty 6.0

    Simple MLPs using temporal and behavioral features from gossip data predict Lightning Network channel closure types better than temporal graph neural networks.

  13. FAME: Forecasting Academic Impact via Continuous-Time Manifold Evolution

    cs.LG 2026-05 unverdicted novelty 6.0

    FAME models scientific topic trajectories in continuous time to forecast paper impact more accurately than LLMs by aligning manuscripts with field momentum in a dynamic latent space.

  14. PRISM: Iterative Cross-Modal Posterior Refinement for Dynamic Text-Attributed Graphs

    cs.LG 2026-05 unverdicted novelty 6.0

    PRISM iteratively transforms semantic priors into behavior-conditioned posteriors via cross-modal refinement to improve representation learning on dynamic text-attributed graphs.

  15. Knowledge Is Not Static: Order-Aware Hypergraph RAG for Language Models

    cs.CL 2026-04 unverdicted novelty 6.0

    OKH-RAG represents knowledge as ordered hyperedges and retrieves coherent interaction sequences via a learned transition model, outperforming permutation-invariant RAG baselines on order-sensitive QA tasks.

  16. BiTA: Bidirectional Gated Recurrent Unit-Transformer Aggregator in a Temporal Graph Network Framework for Alert Prediction in Computer Networks

    cs.LG 2026-04 unverdicted novelty 6.0

    BiTA redesigns temporal aggregation in TGNs by jointly using bidirectional GRU for sequential dependencies and Transformer for long-range context to improve alert prediction accuracy on real network data.

  17. Can LLM Agents Simulate Dynamic Networks? A Case Study on Email Networks with Phishing Synthesis

    cs.SI 2026-03 unverdicted novelty 6.0

    LLM multi-agent systems augmented with data-driven event triggers and Hawkes processes simulate both micro-level interactions and macroscopic topologies in dynamic email networks for realistic phishing synthesis.

  18. Weisfeiler and Leman Follow the Arrow of Time: Expressive Power of Message Passing in Temporal Event Graphs

    cs.LG 2025-05 unverdicted novelty 6.0

    Introduces consistent event graph isomorphism and a temporal Weisfeiler-Leman algorithm to analyze and improve the expressive power of message passing in temporal event graphs.

  19. Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges

    cs.LG 2021-04 accept novelty 6.0

    Geometric deep learning provides a unified mathematical framework based on grids, groups, graphs, geodesics, and gauges to explain and extend neural network architectures by incorporating physical regularities.

  20. A2QTGN: Adaptive Amplitude Quantum-Integrated Temporal Graph Network for Dynamic Link Prediction

    quant-ph 2026-05 unverdicted novelty 5.0

    A2QTGN combines adaptive quantum amplitude encoding with a temporal graph network to improve dynamic link prediction, showing strong results on five benchmark datasets.

  21. Rethinking Fraud Safety Evaluation: Multi-Round Attacks Reveal Safety-Utility Tradeoffs in Graph-Context LLM Defenders

    cs.CR 2026-05 unverdicted novelty 5.0

    Graph-context LLM fraud defenders improve early refusal under replay and adaptive multi-round attacks compared to text baselines but increase benign over-refusal, with the cost localized to how the LLM consumes struct...

  22. Explainable Graph Neural Networks for Interbank Contagion Surveillance: A Regulatory-Aligned Framework for the U.S. Banking Sector

    cs.LG 2026-04 unverdicted novelty 4.0

    ST-GAT applies spatial-temporal graph attention networks to reconstructed interbank graphs from FDIC Call Reports, achieving 0.939 AUPRC for bank distress prediction with explainable feature importance.

  23. AdvSynGNN: Structure-Adaptive Graph Neural Nets via Adversarial Synthesis and Self-Corrective Propagation

    cs.LG 2026-02 unverdicted novelty 4.0

    AdvSynGNN uses multi-resolution structural synthesis, contrastive objectives, an adaptive transformer, and an adversarial propagation engine with residual label correction to improve node-level predictions on challeng...

Reference graph

Works this paper leans on

141 extracted references · 141 canonical work pages · cited by 22 Pith papers · 10 internal anchors

  1. [1]

    NIPS , pages=

    Interaction networks for learning about objects, relations and physics , author=. NIPS , pages=

    work page

  2. [2]

    IEEE Signal Process

    Geometric deep learning: going beyond euclidean data , author=. IEEE Signal Process. Mag. , volume=

    work page

  3. [5]

    and Leskovec, Jure , title =

    Ying, Rex and He, Ruining and Chen, Kaifeng and Eksombatchai, Pong and Hamilton, William L. and Leskovec, Jure , title =. KDD '18 , year =

    work page

  4. [6]

    and Ying, Rex and Leskovec, Jure , title =

    Hamilton, William L. and Ying, Rex and Leskovec, Jure , title =. NIPS , year =

    work page

  5. [7]

    Journal of Machine Learning Research , year =

    Seyed Mehran Kazemi and Rishab Goel and Kshitij Jain and Ivan Kobyzev and Akshay Sethi and Peter Forsyth and Pascal Poupart , title =. Journal of Machine Learning Research , year =

    work page

  6. [8]

    arXiv:1912.11730 , primaryClass=

    Memory Augmented Graph Neural Networks for Sequential Recommendation , author=. arXiv:1912.11730 , primaryClass=. 2019 , eprint=

    work page arXiv 1912

  7. [11]

    ICLR , year=

    Inductive representation learning on temporal graphs , author=. ICLR , year=

    work page

  8. [12]

    CoRR , volume =

    Seyed Mehran Kazemi and Rishab Goel and Sepehr Eghbali and Janahan Ramanan and Jaspreet Sahota and Sanjay Thakur and Stella Wu and Cathal Smyth and Pascal Poupart and Marcus Brubaker , title =. CoRR , volume =. 2019 , url =

    work page 2019

  9. [13]

    Conference on Uncertainty in Artificial Intelligence , keywords =

    Zhang, Jiani and Shi, Xingjian and Xie, Junyuan and Ma, Hao and King, Irwin and Yeung, Dit-Yan , biburl =. Conference on Uncertainty in Artificial Intelligence , keywords =

    work page

  10. [14]

    International Conference on Learning Representations , year=

    Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting , author=. International Conference on Learning Representations , year=

    work page

  11. [15]

    2017 , isbn =

    Baytas, Inci M. and Xiao, Cao and Zhang, Xi and Wang, Fei and Jain, Anil K. and Zhou, Jiayu , title =. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =. 2017 , isbn =. doi:10.1145/3097983.3097997 , abstract =

    work page doi:10.1145/3097983.3097997 2017

  12. [17]

    Graph Attention Networks , booktitle =

    Petar Velickovic and Guillem Cucurull and Arantxa Casanova and Adriana Romero and Pietro Li. Graph Attention Networks , booktitle =

    work page

  13. [18]

    NIPS Workshop on Bayesian Deep Learning , year=

    Variational Graph Auto-Encoders , author=. NIPS Workshop on Bayesian Deep Learning , year=

    work page

  14. [22]

    NIPS , pages =

    Attention is All you Need , author =. NIPS , pages =. 2017 , url =

    work page 2017

  15. [23]

    Liben-Nowell, David and Kleinberg, Jon , title =. J. Am. Soc. Inf. Sci. Technol. , month = may, pages =. 2007 , issue_date =

    work page 2007

  16. [25]

    ICDM , pages=

    Temporal-relational classifiers for prediction in evolving domains , author=. ICDM , pages=. 2008 , organization=

    work page 2008

  17. [26]

    Applied Intelligence , volume=

    Link prediction in dynamic social networks by integrating different types of information , author=. Applied Intelligence , volume=. 2015 , publisher=

    work page 2015

  18. [27]

    Information Sciences , volume=

    An efficient algorithm for link prediction in temporal uncertain social networks , author=. Information Sciences , volume=. 2016 , publisher=

    work page 2016

  19. [28]

    Information Sciences , volume=

    Sampling-based algorithm for link prediction in temporal networks , author=. Information Sciences , volume=. 2016 , publisher=

    work page 2016

  20. [29]

    Procedia Computer Science , volume=

    Link prediction based on common-neighbors for dynamic social network , author=. Procedia Computer Science , volume=. 2016 , publisher=

    work page 2016

  21. [30]

    DEXA , pages=

    A hybrid time-series link prediction framework for large social network , author=. DEXA , pages=. 2012 , organization=

    work page 2012

  22. [31]

    INFORMS Journal on Computing , volume=

    The time-series link prediction problem with applications in communication surveillance , author=. INFORMS Journal on Computing , volume=. 2009 , publisher=

    work page 2009

  23. [32]

    Data Mining and Knowledge Discovery , volume=

    Link prediction using time series of neighborhood-based node similarity scores , author=. Data Mining and Knowledge Discovery , volume=. 2016 , publisher=

    work page 2016

  24. [33]

    VLDB , volume=

    A particle-and-density based evolutionary clustering method for dynamic networks , author=. VLDB , volume=. 2009 , publisher=

    work page 2009

  25. [34]

    , author=

    What to Do Next: Modeling User Behaviors by Time-LSTM. , author=. IJCAI , volume=

    work page

  26. [35]

    ASONAM , pages=

    Evolutionary clustering and analysis of bibliographic networks , author=. ASONAM , pages=. 2011 , organization=

    work page 2011

  27. [36]

    IJCNN , pages=

    Time series based link prediction , author=. IJCNN , pages=. 2012 , organization=

    work page 2012

  28. [37]

    Physica A: Statistical Mechanics and its Applications , volume=

    A novel time series link prediction method: Learning automata approach , author=. Physica A: Statistical Mechanics and its Applications , volume=. 2017 , publisher=

    work page 2017

  29. [39]

    Journal of Web Semantics , volume=

    Embedding models for episodic knowledge graphs , author=. Journal of Web Semantics , volume=. 2019 , publisher=

    work page 2019

  30. [40]

    AAAI , year=

    Dynamic Network Embedding by Modeling Triadic Closure Process , author=. AAAI , year=

    work page

  31. [44]

    2018 IEEE International Conference on Big Data , pages=

    dynnode2vec: Scalable dynamic network embedding , author=. 2018 IEEE International Conference on Big Data , pages=. 2018 , organization=

    work page 2018

  32. [45]

    TKDD , volume=

    Temporal link prediction using matrix and tensor factorizations , author=. TKDD , volume=. 2011 , publisher=

    work page 2011

  33. [46]

    KDD '18 , pages=

    Netwalk: A flexible deep embedding approach for anomaly detection in dynamic networks , author=. KDD '18 , pages=

    work page

  34. [47]

    , author=

    Link Prediction with Spatial and Temporal Consistency in Dynamic Networks. , author=. IJCAI , pages=

    work page

  35. [48]

    AALTD , pages=

    Node classification in dynamic social networks , author=. AALTD , pages=

    work page

  36. [49]

    , author=

    Dynamic Network Embedding: An Extended Approach for Skip-gram based Network Embedding. , author=. IJCAI , pages=

    work page

  37. [50]

    Expert Systems with Applications , volume=

    An adaptive random walk sampling method on dynamic community detection , author=. Expert Systems with Applications , volume=. 2016 , publisher=

    work page 2016

  38. [51]

    Efficient Representation Learning Using Random Walks for Dynamic Graphs

    Efficient representation learning using random walks for dynamic graphs , author=. arXiv:1901.01346 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 1901

  39. [52]

    ASONAM , title=

    S. ASONAM , title=. 2018 , volume=

    work page 2018

  40. [54]

    International Conference on Multimedia Modeling , pages=

    evolve2vec: Learning network representations using temporal unfolding , author=. International Conference on Multimedia Modeling , pages=. 2019 , organization=

    work page 2019

  41. [55]

    G. H. 2018 IEEE International Conference on Big Data , title=. 2018 , volume=

    work page 2018

  42. [57]

    IFAC-PapersOnLine , volume=

    Learning graph dynamics using deep neural networks , author=. IFAC-PapersOnLine , volume=. 2018 , publisher=

    work page 2018

  43. [58]

    Pattern Recognition , volume=

    Dynamic graph convolutional networks , author=. Pattern Recognition , volume=. 2020 , publisher=

    work page 2020

  44. [61]

    WSDM , pages=

    DySAT: Deep Neural Representation Learning on Dynamic Graphs via Self-Attention Networks , author=. WSDM , pages=

    work page

  45. [62]

    2017 , booktitle =

    Trivedi, Rakshit and Dai, Hanjun and Wang, Yichen and Song, Le , title =. 2017 , booktitle =

    work page 2017

  46. [63]

    ICLR , year=

    DyRep: Learning Representations over Dynamic Graphs , author=. ICLR , year=

    work page

  47. [64]

    European Conference on Information Retrieval , pages=

    Relationship prediction in dynamic heterogeneous information networks , author=. European Conference on Information Retrieval , pages=. 2019 , organization=

    work page 2019

  48. [66]

    2019 , journal=

    Temporal Knowledge Graph completion based on time series Gaussian embedding , author=. 2019 , journal=

    work page 2019

  49. [71]

    ECCV , pages=

    Learning human-object interactions by graph parsing neural networks , author=. ECCV , pages=

    work page

  50. [72]

    CVPR , year=

    Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs , author=. CVPR , year=

    work page

  51. [73]

    ICMLA , year=

    Graph neural networks for icecube signal classification , author=. ICMLA , year=

    work page

  52. [74]

    NIPS , year =

    Convolutional Networks on Graphs for Learning Molecular Fingerprints , author =. NIPS , year =

    work page

  53. [75]

    ICML , year =

    Neural Message Passing for Quantum Chemistry , author =. ICML , year =

    work page

  54. [76]

    Med Image Anal , volume=

    Disease prediction using graph convolutional networks: Application to Autism Spectrum Disorder and Alzheimer’s disease , author=. Med Image Anal , volume=. 2018 , publisher=

    work page 2018

  55. [78]

    Bioinformatics , volume=

    Modeling polypharmacy side effects with graph convolutional networks , author=. Bioinformatics , volume=

    work page

  56. [79]

    Scientific Reports , volume=

    HyperFoods: Machine intelligent mapping of cancer-beating molecules in foods , author=. Scientific Reports , volume=

    work page

  57. [80]

    Nature Methods , volume=17, pages=

    Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning , author=. Nature Methods , volume=17, pages=

    work page

  58. [81]

    KDD Workshop on Deep Learning on Graphs , year=

    ncRNA Classification with Graph Convolutional Networks , author=. KDD Workshop on Deep Learning on Graphs , year=

    work page

  59. [82]

    NIPS , year=

    Link prediction based on graph neural networks , author=. NIPS , year=

    work page

  60. [83]

    and Hui, Pik-Mai and Harper, F

    Nguyen, Tien T. and Hui, Pik-Mai and Harper, F. Maxwell and Terveen, Loren and Konstan, Joseph A. , title =. 2014 , isbn =. doi:10.1145/2566486.2568012 , booktitle =

    work page doi:10.1145/2566486.2568012 2014

  61. [84]

    and Welling, Max , biburl =

    Kipf, Thomas N. and Welling, Max , biburl =. ICLR , interhash =

    work page

  62. [85]

    An efficient algorithm for link prediction in temporal uncertain social networks

    Nahla Mohamed Ahmed and Ling Chen. An efficient algorithm for link prediction in temporal uncertain social networks. Information Sciences, 331: 0 120--136, 2016

    work page 2016

  63. [86]

    Sampling-based algorithm for link prediction in temporal networks

    Nahla Mohamed Ahmed, Ling Chen, Yulong Wang, Bin Li, Yun Li, and Wei Liu. Sampling-based algorithm for link prediction in temporal networks. Information Sciences, 374: 0 1--14, 2016

    work page 2016

  64. [87]

    evolve2vec: Learning network representations using temporal unfolding

    Nikolaos Bastas, Theodoros Semertzidis, Apostolos Axenopoulos, and Petros Daras. evolve2vec: Learning network representations using temporal unfolding. In International Conference on Multimedia Modeling, pp.\ 447--458. Springer, 2019

    work page 2019

  65. [88]

    Interaction networks for learning about objects, relations and physics

    Peter W Battaglia, Razvan Pascanu, Matthew Lai, Danilo Jimenez Rezende, et al. Interaction networks for learning about objects, relations and physics. In NIPS, pp.\ 4502--4510, 2016

    work page 2016

  66. [89]

    Relational inductive biases, deep learning, and graph networks

    Peter W Battaglia, Jessica B Hamrick, Victor Bapst, Alvaro Sanchez-Gonzalez, Vinicius Zambaldi, Mateusz Malinowski, Andrea Tacchetti, David Raposo, Adam Santoro, and Ryan Faulkner. Relational inductive biases, deep learning, and graph networks. arXiv:1806.01261, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  67. [90]

    Bronstein, Amra Deli \'c , et al

    Luca Belli, Sofia Ira Ktena, Alykhan Tejani, Alexandre Lung-Yut-Fon, Frank Portman, Xiao Zhu, Yuanpu Xie, Akshay Gupta, Michael M. Bronstein, Amra Deli \'c , et al. Privacy-preserving recommender systems challenge on twitter's home timeline. arXiv:2004.13715, 2020

    work page arXiv 2004

  68. [91]

    Geometric deep learning: going beyond euclidean data

    Michael M Bronstein, Joan Bruna, Yann LeCun, Arthur Szlam, and Pierre Vandergheynst. Geometric deep learning: going beyond euclidean data. IEEE Signal Process. Mag., 34 0 (4): 0 18--42, 2017

    work page 2017

  69. [92]

    Gc-lstm: Graph convolution embedded lstm for dynamic link prediction

    Jinyin Chen, Xuanheng Xu, Yangyang Wu, and Haibin Zheng. Gc-lstm: Graph convolution embedded lstm for dynamic link prediction. arXiv:1812.04206, 2018

    work page arXiv 2018

  70. [93]

    In: Moschitti, A., Pang, B., Daelemans, W

    Kyunghyun Cho, Bart van Merri \"e nboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. Learning phrase representations using RNN encoder -- decoder for statistical machine translation. In EMNLP, pp.\ 1724--1734, 2014. doi:10.3115/v1/D14-1179. URL https://www.aclweb.org/anthology/D14-1179

    work page doi:10.3115/v1/d14-1179 2014

  71. [94]

    Bronstein, Spencer Klein, and Joan Bruna

    Nicholas Choma, Federico Monti, Lisa Gerhardt, Tomasz Palczewski, Zahra Ronaghi, Prabhat Prabhat, Wahid Bhimji, Michael M. Bronstein, Spencer Klein, and Joan Bruna. Graph neural networks for icecube signal classification. In ICMLA, 2018

    work page 2018

  72. [95]

    Time series based link prediction

    Paulo Ricardo da Silva Soares and Ricardo Bastos Cavalcante Prud \^e ncio. Time series based link prediction. In IJCNN, pp.\ 1--7. IEEE, 2012

    work page 2012

  73. [96]

    HyTE: Hyperplane-based temporally aware knowledge graph embedding

    Shib Sankar Dasgupta, Swayambhu Nath Ray, and Partha Talukdar. H y TE : Hyperplane-based temporally aware knowledge graph embedding. In EMNLP, pp.\ 2001--2011, 2018. doi:10.18653/v1/D18-1225. URL https://www.aclweb.org/anthology/D18-1225

    work page doi:10.18653/v1/d18-1225 2001

  74. [97]

    De Winter , T

    S. De Winter , T. Decuypere , S. Mitrović , B. Baesens , and J. De Weerdt . Combining temporal aspects of dynamic networks with node2vec for a more efficient dynamic link prediction. In ASONAM, pp.\ 1234--1241, 2018

    work page 2018

  75. [98]

    Dynamic network embedding: An extended approach for skip-gram based network embedding

    Lun Du, Yun Wang, Guojie Song, Zhicong Lu, and Junshan Wang. Dynamic network embedding: An extended approach for skip-gram based network embedding. In IJCAI, pp.\ 2086--2092, 2018

    work page 2086

  76. [99]

    Temporal link prediction using matrix and tensor factorizations

    Daniel M Dunlavy, Tamara G Kolda, and Evrim Acar. Temporal link prediction using matrix and tensor factorizations. TKDD, 5 0 (2): 0 1--27, 2011

    work page 2011

  77. [100]

    Convolutional networks on graphs for learning molecular fingerprints

    David K Duvenaud, Dougal Maclaurin, Jorge Iparraguirre, Rafael Bombarelli, Timothy Hirzel, Alan Aspuru-Guzik, and Ryan P Adams. Convolutional networks on graphs for learning molecular fingerprints. In NIPS. 2015

    work page 2015

  78. [101]

    Relationship prediction in dynamic heterogeneous information networks

    Amin Milani Fard, Ebrahim Bagheri, and Ke Wang. Relationship prediction in dynamic heterogeneous information networks. In European Conference on Information Retrieval, pp.\ 19--34. Springer, 2019

    work page 2019

  79. [102]

    Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning

    Pablo Gainza et al. Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning. Nature Methods, 17: 0 184--192, 2019

    work page 2019

  80. [103]

    Learning sequence encoders for temporal knowledge graph completion

    Alberto Garc \' a-Dur \'a n, Sebastijan Duman c i \'c , and Mathias Niepert. Learning sequence encoders for temporal knowledge graph completion. In EMNLP, pp.\ 4816--4821, 2018. doi:10.18653/v1/D18-1516

    work page doi:10.18653/v1/d18-1516 2018

Showing first 80 references.