Słownik AI
Kompletny słownik sztucznej inteligencji
Temporal Graph Neural Network (TGNN)
Neural network architecture specifically designed to process graphs where nodes, edges, and attributes dynamically evolve over time.
Time-Aware Message Passing
Information propagation mechanism in TGNNs that integrates temporal information to weight and aggregate messages between neighboring nodes.
Continuous-Time Dynamic Graphs
Graph representation where structural changes are modeled as events occurring at specific moments in a temporal continuum.
Temporal Edge Features
Attributes associated with graph edges that vary over time, including timestamps, temporal weights, or dynamic features.
Temporal Node Embeddings
Vector representations of nodes that evolve dynamically to capture structural and temporal changes in the graph.
Temporal Attention Mechanism
Attention technique that assigns weights to temporal interactions based on their temporal and structural relevance.
Time-Decay Function
Mathematical function used in TGNNs to model the decreasing influence of past events as time progresses.
Snapshot-based TGNN
Approach where the dynamic graph is divided into discrete snapshots in time, each processed by a static GNN.
Event-based TGNN
Architecture that processes graph changes as discrete events, allowing incremental updates of node representations.
Temporal Graph Convolution
Convolution operation adapted for temporal graphs that combines spatial and temporal information during neighborhood aggregation.
Temporal Link Prediction
Task of predicting future connections between nodes in a dynamic graph based on past temporal interactions and evolutions.
Recurrent TGNN
Architecture combining recurrent networks (LSTM/GRU) with GNNs to model temporal dependencies in dynamic graphs.
Temporal Graph Transformer
Variant of the Transformer architecture adapted for temporal graphs with temporal and structural attention mechanisms.
Dynamic Node Classification
Node classification task in evolving graphs where labels can change over time.
Graph ODE Networks
Models using ordinary differential equations to model the continuous evolution of node states over time.
Temporal Random Walk
Walk sampling method in temporal graphs that respects the temporal causality constraint during navigation.
Memory-based TGNN
Architecture using an external memory module to efficiently store and retrieve past temporal information of nodes and interactions.
Temporal Neighborhood Aggregation
Process of aggregating information from temporal neighbors of a node while considering the chronological order of interactions.