> ML_ALGORITHM // TEMPORAL-GRAPH-NETWORKS-TGN_v1.0
Temporal Graph Networks (TGN)
Continuous-time dynamic graph framework that combines recurrent memory modules with graph attention to model streaming timestamped interactions.
Dynamic Graph Neural Networksgraph-relationalmoderate-posthoclarge (>100k)
Back to All AlgorithmsComputational Complexity
Training Complexity:O(events * (memory_update + embedding_generation))
Inference Complexity:O(memory_update)
Hardware Profile
CPU Friendly:No
Requires GPU:Yes
Memory Footprint:high
Interpretability & Data
Interpretability Tier:moderate-posthoc
Training Data Needs:large (>100k)
Interpretability Assessment
Maintains an up-to-date internal state memory vector for every node updated continuously upon each event.
Suitable Tasks & Supported Modalities
Suitable Tasks:
dynamic link predictiondynamic node classificationfinancial fraud detection
Supported Modalities:
graphtime-series
Implementing Libraries
torch-geometric
tgn-official
Foundational Literature
Common Pitfalls & Warnings
- Future interaction leakage if mini-batches are not strictly sorted in chronological timestamp order during training