- Data Observatory (DO), in collaboration with the Millennium Institute of Astrophysics (MAS), and the Center for Mathematical Modelling (CMM), wants this project to position Chile as one of the world’s leading providers of the artificial intelligence layer for the processing of millions of astrophysical alerts per night, and to enable technology transfer in Data Science skills to other industrial and economic sectors.
- ALeRCE is an automatic learning platform for the rapid classification of astronomical events, and it is being prepared to process up to 10 million detections per night through the Vera Rubin Observatory, expected to be launched in 2023.
Data Observatory, along with the Center for Mathematical Modelling (CMM) and the Millennium Institute of Astrophysics (MAS) announced their alliance to promote its Automatic Learning for the Rapid Classification of Events (ALeRCE) Project, and thus position Chile as an international astronomical broker in charge of the artificial intelligence layer that will process the enormous volumes of alerts generated by gigantic telescopes such as the Vera Rubin Observatory, formerly known as the Large Synoptic Survey Telescope or LSST in honor of its Legacy Survey of Space and Time, in a sustainable manner over the next ten years.
Andrés Jordán, scientific leader at Data Observatory and Director of the Millennium Institute of Astrophysics (MAS), says that the initial phase in Data Observatory went from the earth to the sky, dealing with the challenge of collecting and processing the enormous volume of astronomical data and, in this sense, one of the main trials DO is currently facing is related to the avalanche of real time data that will be provided by very large and powerful telescopes and observatories with an extensive scope and capacity such as the LSST. Jordán explains that ALeRCE represents one of the most transversal projects within the astronomical community in Chile, and is crucial for being able to tackle the new astrophysical dilemmas regarding the provision of layers of artificial intelligence to extract the best of science from these machines. This project, currently being promoted by academia, is made up of astrophysicists, statisticians, data scientists, and a great variety of experts and professionals.
“This project is not only important in the Chilean context, since we are preparing to compete and become one of the world´s main brokers of the artificial intelligence layer in charge of millions of astrophysical alerts per night, and in order to do this, we need to be sustainable for at least a decade”, said Jordán.
On the other hand, Alejandro Maass, Director of the Center for Mathematical Modelling (CMM) at Universidad de Chile, mentioned that: “In order to consistently lead a highly competitive international project such as ALeRCE over time, we require to combine the complementary technical efforts of many actors, from mathematics and sophisticated techniques in artificial intelligence to computing power, in order to provide a service that gives added value to the data obtained from astronomy.
Francisco Förster, Principal Investigator at ALeRCE and CMM, as well as Associate Investigator at MAS, explains that the project is an idea that began within the framework of an earlier sky monitoring program carried out during the years 2013 and 2015, where a high capacity camera named the Dark Energy Camera (DECam) searched for supernovae in real time. “This project was the starting point to create an interdisciplinary team of computer engineers, mathematicians, and astronomers that developed tools to observe, ingest and process data with the purpose of discovering supernovae in real time with the use of machine learning tools. Thus, we produced and processed a stream of approximately 100 thousand objects per night. In the future, other telescopes will produce alerts at an even higher rate.”
Carlos Jerez, Executive Director of Data Observatory, and Dean of the Faculty of Engineering and Sciences at Universidad Adolfo Ibáñez, mentioned: “AleRCE is a World Class project that adheres to the efforts of DO to contribute to the country´s scientific and technological development. We hope that this initiative will further translate into learning experiences that will enable the transfer of knowledge and technology applied to innovative productive engines in Chile”.
Until the year 2014, the ALeRCE Project (previously called HiTS or High Cadence Transient Survey), used a 500-megapixel camera with an observation strategy through which the telescope visited the same sky region every 2 hours, while the LSST telescope will have 3000 megapixels and will visit the same sky region typically every 3 days. After the data were received from DECam, they were observed and sent via optical fibre to the National Laboratory for High Performance Computing (NLHPS) in Santiago, and processed in real time. “This exercise allowed us to simulate future scenarios such as the LSST telescope in Chile, scheduled for the year 2023. The LSST will produce 10 million events per night, and we were producing 100 thousand. This is where the question regarding how to know what is to follow appears, classifying and setting priorities according to importance. An astronomical broker is an instrument that connects telescopes in charge of monitoring the sky with others that track interesting objects and require real time monitoring”, Förster points out.
This is an interesting milestone for astronomers, but as a set of data, it represents an opportunity to develop new algorithms, test new structures, and form interdisciplinary teams. Experts have declared this is the only way for the model to be replicated in other industrial and scientific sectors.
Since the LSST will begin to operate in 2023, the team at ALeRCE searched for another observatory to pursue training. They chose the Zwicky Transient Facility (ZTF), a systematic search project for transient short-lived phenomena, lasting from seconds to years, in the Northern Hemisphere, in relation to the astronomical scale that is measured in multiples of millions of years and encompasses phenomena from novae and supernovae to asteroids and comets.
This telescope is located in Monte Palomar, California, and produces 200 thousand alerts per night which ALeRCE has been ingesting for a year. In this sense, Francisco Förster mentions: “The very high international competition we are facing compels us to grow as well as to work within a collaborative network. We require many resources and we trust that Data Observatory is an ally that will help us endure over time, providing us with the necessary conditions to work with the LSST in the future according to its capacity and technical requirements.”
In this alliance with Data Observatory, ALeRCE will be able to begin a new stage of development and expansion and will be upgraded by hybrid infrastructure to guarantee data availability and effective use among the astronomical community. ALeRCE registers over 2000 users in 50 countries, and is expected to increase this impact. The main users are located in Chile, the United States and Japan.
“Human capital is much more important than the infrastructure, since in the short term infrastructure can become obsolete whereas people can be further educated and adapt to new technologies”, Förster assures. Open databases are an experimental field for preparing human capital. From the point of view of transference, if data lack a group of people that knows how to process them, they are useless. In Chile, we have human capital that already knows how to do the work and can boost our investigation and economy today”, Förster concludes.
Regarding ALeRCE indicators, in one year it has processed over 100 million alerts in real time, including 20 million images. It has reported more than 3000 supernovae, 60 thousand supermassive black holes or 800 thousand variable stars. More information at: http://alerce.science/. Video of the project: http://alerce.science/