KI-Glossar
Das vollständige Wörterbuch der Künstlichen Intelligenz
Contextual Anomaly
An observation that is statistically normal in isolation but becomes abnormal when considered within its specific temporal or event context.
1D Convolutional Auto-encoder (1D-CNN)
A neural network architecture using one-dimensional convolutions to learn a compressed representation of time series, facilitating reconstruction and error detection.
Isolation Forest
An unsupervised learning algorithm that isolates anomalies by building random decision trees, where outliers are easier to isolate and require fewer partitions.
Isomap for Time Series
A nonlinear dimensionality reduction technique that preserves geodesic distances, used to visualize and detect anomalies in high-dimensional spaces.
LSTM Auto-encoder
A recurrent neural network model using LSTM units to learn the temporal structure of a series, where a high reconstruction error signals an anomaly.
Exponentially Weighted Moving Average (EWMA)
A smoothing method that assigns exponentially decreasing weights to past observations, serving as a basis for adaptive anomaly detection thresholds.
Quantile Mapping
A statistical technique that transforms a data distribution to match a reference distribution, used to normalize data before anomaly detection.
Local Outlier Factor (LOF)
A measure of the local density of a data point relative to its neighbors, where a high score indicates that the point lies in a region of low density, suggesting an anomaly.
TSFresh
A Python library that automatically extracts a large number of temporal features, which can be used to train supervised anomaly detection models.
Learning Vector Quantization (LVQ)
A supervised learning algorithm that uses prototypes for classification, adaptable for anomaly detection by identifying observations distant from the prototypes.