A framework for using LSTMs to detect anomalies in multivariate time series data. Includes spacecraft anomaly data and experiments from the Mars Science Laboratory and SMAP missions.
Can this algorithm be extended to determine different types of anomalies? For example, different control inputs and disturbances to a system can produce different classes of anomalous data (anomaly due to high operating temperature, short circuit, excessive load etc). How do I extend the network to determine which control input and disturbance led to the particular anomaly in order to classify the type of anomaly? From my understanding the network currently only determines if there is an anomaly or not.
Can this algorithm be extended to determine different types of anomalies? For example, different control inputs and disturbances to a system can produce different classes of anomalous data (anomaly due to high operating temperature, short circuit, excessive load etc). How do I extend the network to determine which control input and disturbance led to the particular anomaly in order to classify the type of anomaly? From my understanding the network currently only determines if there is an anomaly or not.