Linguae-ML Seminar
Machine Learning on Temporal Graphs
From Foundations to Emerging Frontiers
Networks are everywhere
And graphs are a great way to model them
Functional Networks
Interaction Networks
Social Networks
Molecules
Image Credit: wolfram.com · gatton.uky.edu · Papo et al., Frontiers in Human Neuroscience 2014 · Madhavicmu / Wikimedia Commons CC-BY-SA-4.0
Networks are everywhere
And graphs are a great way to model them
Syntax Trees
Cognitive Maps
What model should we use for graphs?
Modality
Data
Architecture
🖼️
Images
📝
Text
The
cat
sat
on
the
mat
🕸️
Graphs
?
Graph Neural Networks (GNNs)
Convolutional GNN
$\mathbf{H}^{(k)}$
=
$\sigma(\tilde{\mathbf{A}} \mathbf{H}^{(k-1)} \mathbf{W}^{(k)})$
$\mathbf{H}^{(0)}$
=
$\mathbf{X}$
Message-Passing GNN
$\mathbf{m}_i^{(k)}$
=
$\text{AGG}^{(k)}\!\left(\{\{\mathbf{h}_j^{(k-1)}: (i,j) \in E\}\}\right)$
$\mathbf{h}_i^{(k)}$
=
$\text{COM}^{(k)}\!\left(\mathbf{h}_i^{(k-1)}, \mathbf{m}_i^{(k)}\right)$
$\mathbf{h}_i^{(0)}$
=
$\mathbf{x}_i$
Some Examples of Dynamic Graphs
Graphs changing over time
Social Networks
Interaction Networks
From Static to Dynamic Graphs
$G=(V,E,X)$
Static Graph
• No notion of time
• Examples: molecules, syntax trees
• Examples: molecules, syntax trees
$G_t=(V,E,X_t)$
Spatio-Temporal Graph
• Fixed topology; changing features
• Regular time intervals
• Examples: traffic, weather sensors
• Regular time intervals
• Examples: traffic, weather sensors
$G(t)=\{x_{t_1},x_{t_2},...\}\quad t_1\leq t_2\leq ...$
Continuous-Time DG (CTDG)
• Most general formulation
• Sequence of timestamped events
• Examples: social, financial
• Sequence of timestamped events
• Examples: social, financial
$G_t=(V_t,E_t,X_t)$
Discrete-Time DG (DTDG)
• Changing topology and features
• Regular time intervals -> sequence of snapshots
• Examples: weekly trade networks
• Regular time intervals -> sequence of snapshots
• Examples: weekly trade networks
Less General
More General
Why Learning on Dynamic Graphs is Different
Temporal graph models must exploit the event history without collapsing it into one static snapshot
What changes
- Events arrive in order: creation, deletion, update, interaction
- The model should use both what happened and when it happened
- New tasks ask what happens next, or when it happens
Why static GNNs are not enough
event history
G(t ≤ T)
G(t ≤ T)
→
latest snapshot only
- Information loss: the last snapshot hides the evolution path
- Inefficiency: every new event can force repeated computation
- No timing prediction: static GNNs do not support predicting when something will happen
TGN: Temporal Graph Networks
Our work: a modular framework that combines memory-based and graph-based temporal learning
TGN contribution
- General framework combining the best of memory-based and graph-based approaches
- Updates node memories from events, then uses a GNN over the interaction graph to produce embeddings
- Generalizes previous memory-based models and graph-based models in one notation
TGN: State-of-the-Art in 2020
High accuracy with much lower per-epoch cost than the previous methods
TGN: Still a Strong Baseline Today
Years of follow-up work, yet TGN remains hard to beat on standard benchmarks
Temporal Graph Benchmark
Diverse, large datasets and unified evaluation
Key Findings
- Previous datasets were too easy and saturated
- Exposed that simple historical baselines can be surprisingly strong
- Large new datasets make scalability part of the benchmark
9 datasets · 5 domains · up to 72M edges
Temporal Graph Benchmark 2.0
TGB but for knowledge graphs
Key findings
- Edge and relation type information is crucial for strong performance
- Simple heuristic baselines remain competitive with more complex methods
- Many methods fail to run on the largest datasets, making scalability a central result
8 datasets · 5 domains · up to 53M edges
UTG: Unifying Snapshot & Event-Based Models
Huang, Poursafaei, Rabbany, Rabusseau, Rossi, LoG 2024
The problem
Snapshot (DTDG) and event-based (CTDG) models developed in isolation: limited cross-comparison and no unified evaluation.
Contribution
- Input mapper: convert CTDG to snapshots and DTDG to events, so any model can run on any data
- UTG training: use streaming training to make snapshot-based models operate on event streams
- Output mapper: align predictions to either continuous-time or discrete-time tasks
Foundation Models for TG
Existing TG models are trained and tested on the same graph.
Can we train a single TG model that works on unseen graphs, ideally from different domains?
MiNT: Multi-Network Transfer Benchmark for Temporal Graph Learning
- New dataset: 84 distinct ERC-20 token transaction networks
- New framework to train existing TG models on multiple graphs simultaneously
- Scaling law for TGs: model performance improves as it is trained on more networks
Can we use LMs directly on temporal graphs?
TGTalker
Translate temporal graph structure into natural language and feed it to a pre-trained LLM — no fine-tuning required.
Four prompt components
- Background set — recent edges as context
- Example set — 5-shot Q&A pairs
- Query set — target edge to predict
- Temporal neighbors — 1-hop recent neighbours
TGTalker: Results & Explainability
Results
- Competitive without fine-tuning
- Consistently outperforms TGN and HTGN
10 explanation categories discovered
- Most Recent Interaction ≈ EdgeBank heuristic
- Most Frequent Destination ≈ PopTrack heuristic
- Novel patterns: sequence logic, analogy-based inference
🔊 Communicating Sound Through Natural Language
A sender agent describes a sound; a receiver agent decodes the description back to audio
| Method |
Human readability
|
LLM-native transport
|
Semantic editing
|
Acoustic interpretability
|
Training-free
|
Generative decoding
|
Bandwidth efficiency
|
|
|---|---|---|---|---|---|---|---|---|
| Lossless codec (FLAC, WAV) | ||||||||
| Handcrafted descriptors (MFCC, spectral centroid) | ||||||||
| Neural codec (EnCodec, SoundStream) | ||||||||
| Audio-language tokenizers (AudioLM, TASTE) | ||||||||
| Unconstrained text caption | ||||||||
| LAC (this paper) |