Add ability to stop the learning phase for the detector

[tejpochiraju]

Is your feature request related to a problem? Please describe. Currently, the detector continues learning from new data - even if this is anomalous. This results in unexpected behaviour such as abnormal data being normalised if a failure condition persists. This is often the case with e.g. industrial sensors.

Imagine a temperature sensor in a process with normal temperatures between 30C-80C. In a failure, the sensor detects 100C and the failure persists. In such a case the anomaly detector will initially detect the anomaly and gradually lower its grade until 100C is no longer considered anomalous by the detector.

Describe the solution you'd like Give users the ability to pause or stop training of the detector so that new data, anomalous or otherwise, is not used for training.

Describe alternatives you've considered Rule based detectors can catch these use cases but are cumbersome.

Additional context I have written a blog article describing the overall flow and use case.

[kaituo]

Thanks Tej. I will discuss with the team and reply later.

[kaituo]

Hi Tej, we accepted the feature request and will put this to our backlog. You are also welcome to contribute the changes if you have bandwidth.