Glossario IA
Il dizionario completo dell'Intelligenza Artificiale
Community Outlier Detection
Method for identifying nodes or groups that do not logically integrate into existing graph communities, revealing isolated or malicious entities.
Autoencoder for Graph Anomaly Detection
Unsupervised learning model that reconstructs graph features, with anomalies identified by high reconstruction error indicating significant deviation.
One-Class SVM on Graphs
Semi-supervised learning algorithm that learns a decision boundary around normal graph data, classifying points outside this boundary as anomalies.
Random Walk Based Anomaly Detection
Approach using random walks to explore graph structure and identify regions or nodes with abnormally low or high transition probabilities.
Temporal Graph Anomaly Detection
Specialized technique for detecting anomalies in evolving graphs, considering structural and behavioral changes over time.
Graphlet-Based Anomaly Detection
Method analyzing the frequency of small graph patterns (graphlets) to identify regions with abnormal pattern distributions, revealing suspicious behaviors.
Attribute Anomaly Detection in Graphs
Anomaly detection based on node or edge attribute values, combining structural and semantic information for more accurate identification.
Structural Anomaly Detection
Approach focusing exclusively on graph topology to identify unusual structures without considering node or edge attributes.
Label Propagation for Anomaly Detection
Semi-supervised algorithm propagating labels through the graph to identify abnormal nodes based on inconsistency with their structural neighbors.
Graph Convolutional Networks for Anomaly Detection
Variant of GCNs specialized in learning anomaly-sensitive representations for effective detection in large-scale complex graphs.
Adversarial Graph Anomaly Detection
Detection framework using adversarial techniques to improve model robustness against sophisticated anomalies and graph attacks.
Deep Graph Infomax for Anomaly Detection
Representation learning method maximizing mutual information between global and local graph representations for robust anomaly detection.