Anomaly Detection Annotator

Support Lasair
#1

Hi,

I was looking to integrate some research I’ve done on anomaly detection ([2403.14742] A Classifier-Based Approach to Multi-Class Anomaly Detection for Astronomical Transients) into the Lasair alert stream.

I currently have an script that runs periodically to check for new alerts, and locally calculates anomaly scores and classification outputs.

How can I go about integrating this as an annotator?

Thanks,
Rithwik

#2

Hi Rithwik
The annotation pattern is sketched below. You might want to start first with Lasair-ZTF, since there will be no LSST alerts for a few more weeks.

  • Make a pre-filter using the Lasair features, see here and here. It could be based on host galaxy, your own watchlist, or any of the Lasair schema, see here.
  • Make your filter active, so that it runs in real time, not just when you click on it. It will then have a Kafka output stream. Your filter description will say something like " The filter is streamed via kafka with the topic name lasair_123blahbla
  • Use the topic name to build a code that receives the alerts, see here and here. Your code can fetch the full lightcurve using the API, then build its annotation.
  • Ask the Lasair team to make you an annotator, see here.
  • Run your code whenever you like to pull the latest alerts and analyse them.
    Cheers
    Roy