Słownik AI
Kompletny słownik sztucznej inteligencji
Stream Clustering
A set of techniques aimed at partitioning data that arrives continuously and potentially infinitely, in real-time and with limited resources.
Macro-cluster
A stable, long-term representation of a cluster, often derived from the merging or evolution of micro-clusters to capture persistent trends in the stream.
Landmark Window
A memory model that processes all data from a fixed starting point in time, useful for analyzing evolution since a landmark event.
Density-Based Stream Clustering
A clustering approach that identifies dense regions of data points in a stream, capable of handling arbitrarily shaped clusters and detecting noise.
DBSTREAM Algorithm
A density-based stream clustering algorithm that uses dense grids and micro-clusters for efficient memory management and rapid drift detection.
Density Factor
A metric used in some stream clustering algorithms to evaluate the density of a micro-cluster, influencing its creation, merging, or removal.
Decaying Weight
A mechanism that assigns decreasing importance to older data points, allowing the model to focus on recent trends in the stream.
Online Clustering
A phase of the process where each new data point is processed and assigned to a micro-cluster incrementally, without requiring the entire dataset.
Offline Clustering
Optional phase that generates the final macro-clusters from existing micro-clusters, often on user demand for analysis at a specific point in time.
Dynamic Grid
Spatial data structure that adapts by dividing or merging cells to track the evolution of data distribution in a stream, optimizing memory usage.
Stream Anomaly Detection
Process integrated into stream clustering that identifies data points not belonging to any dense cluster, flagging them as anomalies or noise.
Cluster Synopsis
Compact representation of a cluster (or a micro-cluster) containing essential statistics such as center, radius, and weight, enabling efficient calculations.
DenStream Algorithm
Density-based stream clustering algorithm that distinguishes potential micro-clusters from core micro-clusters to model emerging and stable clusters.
Time Horizon
Parameter defining the relevance period of data in a stream clustering model, influencing the speed at which the model forgets old information.