VaelstmDetector
Detect anomalies in time series using combined variational autoencoder (VAE) and long short-term memory (LSTM) networks
Since R2025a
Description
The VaelstmDetector object uses a deep learning network
architecture that contains both VAE and LSTM networks to implement a
detector model capable of being trained to detect anomalies in time
series data using only normal data. You create this object with the vaelstmAD
function.
By comparing forecasted values with measured data within a detection window, the detector identifies anomalies as significant differences between forecasted and measured observations. You can control the sensitivity of the detector by modifying a set of threshold and window-sizing properties.
Creating a VaelstmDetector object is the first step in a workflow that
includes creation, training, testing, assessing, and, if necessary, modifying the detector.
For information on the workflow for developing a Predictive Maintenance Toolbox™ anomaly detector, see Detecting Anomalies in Time Series Using Deep Learning Detector Models.
This anomaly detector model was inspired by the architecture proposed in the paper in [1] .
For more information on the functions this workflow uses, see Object Functions.
Creation
Create a VaelstmDetector object by using the vaelstmAD
function.
Properties
Object Functions
train | Train anomaly detector and obtain detection threshold |
detect | Detect anomalies in time series using trained detector model |
plot | Plot detected anomalies and anomaly scores in time series |
plotHistogram | Plot histogram of anomaly scores and detection threshold |
updateDetector | Update settings of a trained anomaly detector and recompute detection threshold |
References
[1] Lin, Shuyu, Ronald Clark, Robert Birke, Sandro Schonborn, Niki Trigoni, and Stephen Roberts. “Anomaly Detection for Time Series Using VAE-LSTM Hybrid Model.” In ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 4322–26. Barcelona, Spain: IEEE, 2020. https://doi.org/10.1109/ICASSP40776.2020.9053558.
Version History
Introduced in R2025a
