Fink broker: fast and efficient framework for machine learning with LSST

Emille Ishida 1 , Julien Peloton 2 , Anais Moller 3

  • 1 CNRS/LPC Clermont, Clermont Ferrand
  • 2 CNRS/IJCLab, Orsay
  • 3 Swinburne University Centre for Astrophysics and Supercomputing, Melbourne

Abstract

Next generation experiments such as the Vera Rubin Observatory Legacy Survey of Space and Time (LSST) will provide an unprecedented volume of time-domain data opening a new era of big data in astronomy. To fully harness the power of these surveys, we require analysis methods capable of dealing with large data volumes that can identify promising transients within minutes for follow-up coordination. In this talk I will present Fink, a broker developed to face these challenges. Fink is based on distributed computing technology and designed for fast and efficient analysis of big data streams. It has been chosen as one of the official LSST and will provide direct public access to the nightly alert stream. I will highlight the state-of-the-art machine learning techniques used to generate early classification scores for a variety of time-domain phenomena including kilonovae and supernovae, as well as for artifacts, like satellites glitches. Such methods include Deep Learning advances and Active Learning approaches to coherently incorporate available information, delivering increasingly more accurate added values throughout the duration of the survey. I will also highlight the potential for discovery of new category of sources and how we can optimize for discovery in the era of LSST.

Transient Astronomy in the era of big data

The Vera Rubin Observatory Legacy Survey of Space and Time (LSST) is expected to detected 10 million alerts per night for a total run of 10 years. 

Fink is one of the community brokers chosen to receive this full alert stream, filter, add value and redistribute a variety of sub-streams to the astronomical community. 

 

Design and data access

Fink was specifically design to meet the challenges presented by LSST. 

The broker is currently working with data from the Zwicky Transint Facility (ZTF), considered a precursor of LSST (whose operations are expected to start in 2024) .

The main entry point for users is the Science Portal , which allows fast and informative visual inspection of alerts. 

 

See below an example: https://fink-portal.org/ZTF22aaihxzh

Deep and Adaptative machine learning

Fink has a modular structure which enables the inclusion of complex requestes in the form of science modules.

As an example, the Early Supernova Ia module is a combination of Bayesiann deep learning (Moller and Boissiere, 2020) and adaptative learning techniques (Leoni et al., 2022). 

Fink Early Supernova Ia candidates have been reported to the Transient Name Server (TNS) between November/2020 - March/2022:

 

  • 4 661 Early SN Ia candidates
    • 573 spectroscopically classified
    • Contaminants are mostly other SNe
    • 1 LBV

True class of spectroscopically confirmed alerts.

A place for all transient astronomers to interact with the LSST alert stream

Beyond identifying SN-like transients, Fink also has already running modules for kilonovae (see Aivazyan et al., 2022 and the talk by Biswajit Biswas at SS14b), variable stars, microlensing, solar system objects, satellite glints (Karpov and Peloton, 2022) and GRB afterglows.

 

Others are currently being implemeted and we are open to further requests. 

If you have a particular astronomical transient you would like identify within LSST data, do not hesitate to contact us at contact@fink-broker.org. 

 

 

Publications using Fink :

Moller et al., 2020, Fink, a new generation of broker for the LSST community

Leoni et al., 2022, Fink: early supernovae Ia classification using active learning

Aivazyan et al., 2022, GRANDMA Observations of ZTF/Fink Transients during Summer 2021

Karpov and Peloton, 2022, Impact of satellite glints on the transient science on ZTF scale

 

Acknowledgements

This work was developed within the Fink community and made use of the Fink  community broker resources. Fink is supported by LSST-France and CNRS/IN2P3.