Model Metrics
Metrics computed from the most recent training run
--%
Novelty Score
Not present in training data
--%
Uniqueness Score
Distinct generated triples
--%
Train Overlap
Exact matches with training triples
--%
Relation Diversity
Used relation types vs schema
--%
Schema Validity
Valid under the schema
--
Generated RDF Candidates
Produced in the latest run
Relation Distribution
Distribution of generated triples across relation types
Average embedding distance: --
Generated RDF Candidates
The following triples were generated by the model. See SemanticGAN/generated.txt for the full raw dataset.
| Subject | Predicate | Object | Score | Ontology |
|---|---|---|---|---|
| Loading model outputs... | ||||
Training Loss
Training loss across recent epochs
Latent Space Manifold
t-SNE projection of learned entity embeddings in 2D