AI-ordlista
Den kompletta ordlistan över AI
TransE
Translation-based embedding model that represents relations as translation operations between entity vectors in the embedding space.
RotatE
Embedding model using rotations in the complex plane to model different types of symmetric, antisymmetric, and inverse relations.
ComplEx
Embedding model in the complex space capturing asymmetric relations through complex multiplication of entity vectors.
DistMult
Simple bilinear model using the diagonal dot product to predict the plausibility of triplets in a knowledge graph.
RESCAL
Tensorial model factorizing the relation tensor to learn entity embeddings and a relation matrix per relation type.
ConvE
Embedding model based on 2D convolutional neural networks capturing complex interactions between entities and relations.
KBAT
Transformer architecture with attention mechanism adapted for knowledge graphs, integrating structural information and semantic content.
Metapath2Vec
Representation learning algorithm for heterogeneous graphs using metapaths to guide random walks and capture relational semantics.
Poincaré Embedding
Embedding technique in Poincaré hyperbolic space that naturally preserves hierarchical structures of knowledge graphs.
Link Prediction in KG
Process of predicting the existence or plausibility of relationships between entities based on vector similarities in the embedding space.
Entity Alignment
Technique that identifies and matches equivalent entities between different knowledge graphs using their vector embeddings.
Temporal Knowledge Graph Embedding
Extension of embedding methods that incorporates the temporal dimension to model the dynamic evolution of relationships over time.
Type-aware Embedding
Embedding approach that integrates entity type information as constraints to improve the quality of vector representations.
Hole
Circular embedding model using circular convolution (correlation) to capture complex interactions between entities and relationships.
SimplE
Simple yet effective embedding model that uses inverse embeddings to capture the inversion and composition properties of relationships.
TuckER
Embedding model based on TuckER tensor decomposition that captures complex interactions between entities and relationships through a core tensor.
CrossE
Model introducing interaction vectors to capture cross-dependencies between entities and relations in knowledge graphs.