A couple of years ago I had already the opportunity to tell you about the Astroinformatics conference. As stated by the name, this international conference is devoted to this new science that, in the recent year,s is revolutionizing astronomical research, as well as other fields.
In fact, we have nowadays a big problem in astronomy. That is to say, the amount of data available is truly exploding, thanks to the new instruments, like new generation telescopes, detectors and automatic spaceships. The collection and analysis of this data requires the adoption of new techniques, based on the intensive use of machine learning, artificial intelligence, virtual reality and other innovative technologies. From a certain point of view, this is a complete reversal of the way astronomical research is performed: no longer are we trying to collect data in order to prove a certain hypothesis, but instead mining into data to find new correlations and unknown trends.
Thanks to astroinformatics, a new world of discoveries is being unveiled, one which involves all the fields at the frontier of astronomy, from extrasolar planets research to gravitational waves, and studies involving the early Universe, galaxies formation and evolution and so on. In recent articles, just to name some examples, we already spoke about the discoveries made using data from the Gaia and Kepler missions concerning extrasolar planets.
Going back to the conference, it is hosted every year in a different location around the world and this time we had the privilege and burden to organize it in Heidelberg, at the institute where I am currently working, the Heidelberg Institute for Theoretical Studies (HITS).
Apart from the pain and stress of organizing such an event, it has been a fantastic experience, and I would like to inform you about it, through the words of some of the people that were directly involved. Therefore, I interviewed three researchers, with very different roles and careers, to give you an insight about this conference, their feelings and their work concerning astroinformatics and astronomy.
The first one is Dalya Baron, from Israel. She is a very young and talented scientist participating in Astroinformatics conference for the second time. The second researcher, George Djorgovski, is from the USA and one of founders and foremost experts in the field of astroinformatics. Finally, the last interview is Kai Polsterer, from Germany. He is the leader of one of the few groups in the world explicitly dedicated to astroinformatics and he is the main organizer Astroinformatics 2018. In addition, he is also my supervisor. Therefore I am particularly emotionally involved in his interview.
Let's stop here with the introduction then and leave space to what scientists have to say!
Interviews
Could you tell us a few words about the role of astroinformatics with respect to astronomy nowadays? Why do you think this field of research is important?
Dalya: I consider astroinformatics as a way to translate information, methods and important tools developed in other fields, like computer science and statistics, to astronomy. I believe that in the future astroinformatics will become astronomy. The data size and complexity is becoming too large and all of us will need to become an expert in astroinformatics, because without these tools we will not be able to do our science anymore.
George: Astronomy, like all other fields of science, has been transformed by the Big Data avalanche and we need tools to extract knowledge from this data. This is what astroinformatics is all about. It is really meant to help astronomers to extract useful knowledge from the complicated and huge amount of data that we have.
Kai: Like all sciences, the driving force behind astronomy is always in observation and data. Theories are connected with what is observed, but for astronomy this is a little bit more complicated, as you cannot go to a laboratory and do experiments. Astronomy however developed a lot, thanks to computers, and now we can automatically control telescopes, have new and more efficient detectors, with adaptive optics systems and so on. Therefore we are so quick in collecting data and we are quicker and quicker, driven by Moore's law, that the data rate and their complexity is increasing tremendously. Astroinformatics is a way to deal with such complexity, more than just applying machine learning or visualization techniques. It is covering all the aspects for which data analysis is not possible anymore with traditional techniques and demands highly sophisticated and new technologies. Astroinformatics is then at the forefront of this development. We are demanded to come with new trends, new ideas and methods to deal with such an avalanche of data.
The Astroinformatics conference series has been going on for several years, starting from a small community and it grew up gathering together expertise from different fields, not just astronomy related. How do you expect this conference will continue to develop and evolve in the future?
Dalya: This is just my second years at this conference and I am only doing astrophysics for few years, but I can already perceive the change. Looking, for example, at the papers coming out on arXiv about machine learning algorithms or statistical methods, I can see that the interest is growing, because it's becoming less and less possible to do things in the traditional way. I would think that this community will become larger and larger, integrating people from different fields and finding a common ground. I also expect that in several years astroinformatics will become a specific course in undergraduate or graduate astronomy, as it will become extremely important to teach students how to work with these tools. Exactly how nowadays we teach them how to use telescopes.
George: More and more people are becoming aware of the need for this type of skills, tools and methods. We see a growing interest in the community, so I only expect this conference will grow in the attendance and there will be more of them.
Kai: It is not about what I expect but more about what I wish to see. I really see a lot of things in astronomy that are really good compared to other natural sciences. We have good standards and a fair way of sharing our data. We do not have to deal too much with privacy issues. Instead what I would really like to see from the Astroinformatics conference is getting more in contact and engaging strong collaborations with other research fields. I would like also seeing a growing younger community more open to accept those new technologies, willing to try new stuff and not just bound to old ways of data analysis and methodologies. This is why I wanted to have an hackathon at this year's conference, to have people playing around and coming with new ideas.
What are your main research interests?
Dalya: I see myself as someone in the middle of different fields. I am very interested in the study of supermassive blackholes in the center of massive galaxies. I am interested in how they affect the host galaxies, how they grow and accrete matter and how this affects the gas in the galaxy and the stellar formation processes due to the black hole activity. The other part of my research is devoted to machine learning algorithm and in particular unsupervised learning. I am interested in understanding how we can do new science with these algorithms and I would like to see if they can teach us things that we are not able to find. And obviously the most interesting thing for me is trying to connect these two fields and see if such algorithms can teach me something about supermassive blackholes and galaxies, without my intervention.
George: I spend my time divided between astronomy, Big Data, astroinformatics, and data science more generally. I also work and cooperate with biologist and geophysicist, or even people from other different fields.
Kai: Off course astroinformatics (laughs). It is hard to say, I am always driven by curiosity. I want to understand new things, find new things. My research interests really range from new methodologies to visualize data to automatic data analysis by using machine learning. I am also interested in data storage problems, how to efficiently access data, including uncertainties in data on database levels. So I have not a specific big question or interest, but I am more interested in developing methodologies to allow answering the big questions. Clearly I am personally also very interested in formation processes in the early universe and other cosmological big questions, understanding dark matter and so on. This is why I was, for example, always interested in photometric redshifts, as I think this is an essential tool to contribute on the comprehension of these big topics.
What makes your work different from other researchers?
Dalya: If I look at myself and many other people here at the conference, I realize that we have to work very hard to convince people from both of these interacting fields that we are doing a good job. Astronomers sometimes look at us suspiciously and this is fine because if you work with tools that look as black boxes, scientists will not trust you. So we are trying to make them trust us and to trust the algorithms we use as well. On the other hand you have the computer scientists community that pretends us to be rigorous in the treatment of data analysis and models. We are trying to take both these aspects and integrate them in order to use them for doing new science. Therefore I always feel the need to learn much more from both fields in order to be able to build something meaningful. For this reason every day for me is different, one day I am more a computer scientist, one day more a physicist and this is funny and very interesting.
George: I think to be among those people most convinced that this is the direction to go. I believe that we need to apply technologies like machine learning, artificial intelligence, virtual reality, and everything we can get from applied computer science and engineering, for the purpose of finding new things in the astronomical data. This is a general problem, not just related to astronomy and we are developing new scientific methodologies for the 21st century. Therefore my main motivation is understanding how computing is changing the way we do science, along with everything else.
Kai: I think we try not to be too biased. We often fail trying new ideas. I accept the risk of failing to try something that no one has tried before, hoping to discover novel approaches and methods. This is what makes our work different. Too often people are trying to get a tiny improvement in this or that single parameter's calculation. It is also important, sure, but I am more interested in some bigger scope, like finding completely new methodologies and standards breaking the previous ones. This includes constant failing and sometimes testing stupid ideas. Then it could happen that sometimes people realize that those stupid ideas were not that stupid in the end. It is funny and it is what I like most about our research.
In recent years there were many important new astronomical discoveries. Which made you most excited, and why?
Dalya: I think most of the people in the community will agree that the LIGO detection of gravitational waves is the most exciting one. Apart from giving another prove to the theory of general relativity, it gives us the possibility to explore a completely new parameter space. It is like developing a new sense to look at Universe. Through them, we can truly feel the Universe. This will allow us to see things which are completely different, completely new, like what happened with neutrinos, or gamma ray astronomy. I think the same will happen with LIGO and I really hope we will soon start finding completely new classes of phenomena.
George: I think Gaia data releases and all the scientific results that came out from it are spectacular. It is really a qualitative jump in our understanding of stellar physics and the Milky Way galaxy. So it is not a single result but many, and it was enabled again by the Big Data. I also have to say that with our sky surveys, like the Catalina Real-Time Transient survey (CRTS), we assembled unprecedented datasets in terms of number of objects, length of coverage and depth. Among many other things it enabled us to do many systematics studies of variability of quasars, for examples. In this field we discovered many new things or improved things that very already known but with much smaller details. The one I like most is the evidence that some quasars contain supermassive binary black holes, exactly as it is predicted by theory, and there is no other way to prove it except that with using periodic variability. The trick was using modern machine learning and statistic tools applied on large datasets. This is a very good example of what astroinformatics is all about: you need data and you need tools.
Kai: Really good question. I must say I was most excited about the discovery of the accelerated expansion of the universe. This is because I grew up in a time where the cosmological picture was completely different. The favorite option for the evolution of the Universe was probably still that of a Big Crunch. Therefore I really thought: "Oh cool, again we came up with something completely different". I think this is one of the coolest things in astronomy, that we constantly learn that everything we thought we understood is completely wrong, or at least needs a strong correction. This tells us that we do not understand at all the universe. Dark matter was already something hard, but with this...
The sixth question is different for every interviewed.
Dalya: Which are the future plans you have for your career?
Hopefully I will stay in academia, which I think is a very nice place and environment. I think academia is the only place where people go working not for money. My supervisor is 74, retired for 10 years, and he still comes to my office every day and looks at my results like an excited child. I want to do the same, I want to stay in academia, be a professor, teach students how amazing science and physics are. This is what I want to do for the rest of my life.
George: When did you hear for the first time the word astroinformatics? Which are the differences between astronomy as it was at the beginning of your career and now?
I don't remember exactly, but it is something that came out quite naturally and started to be used in astronomy in the decade between 2000 and 2010. Then, in 2010 we had the first, inaugural Astroinformatics conference. Speaking of my career, when I started astronomy digital detectors like CCD were just coming in use. Before we were using only photographic plates. Therefore my generation was the first one that had to develop image processing tools for astronomy. Back then a megabyte was a huge data size for a file. In the mid 90s we recognized that information technology evolution is affecting astronomy like every other field and Moore's law is acting in terms of data growth and the ability to process the data. Therefore my scientific interests moved to things that you do with very large telescopes, like studying very distant galaxies, to thing that you do with very huge amount of data. Big sky surveys started to be released, then the virtual observatory framework came up and finally astroinformatics. We have been among the first groups using machine learning to analyze astronomical data and working together with computer scientists. It was obvious to me that this is the future. I've been working on many kinds of different things but I realize that computing is changing all sciences in a qualitative and quantitative fashion and it is exciting to be able to do things that we could have never done before because we simply did not have the data.
Kai: You are the leader of one of the few groups in the world explicitly dedicated to astroinformatics. What is your experience about leading such a group? How do you see the future of this research field?
I would say that it is not about leading a group but more trying to put together different intelligences. In such interdisciplinary field you are never an expert in the whole field. So you need experts from all sides talking to each other to create a heterogenous group, being able to follow individual instincts and trying to balance the individual skills. I see myself more like a shepard, but I am following the sheeps and see where they go, being sure that they stay together on their path.
Concerning the field, I am really frightened by the fact that everyone says that we need more data scientists, politicians, industry and so on. But what we really need in academia is more professors in computer science which work closer with departments of natural sciences, or even other fields. I am afraid that academia, with such tremendous demand from industry, cannot compete, as we cannot pay the same amount of money. So many young scientists, after completing their training and becoming really good researchers, leave for industry. The risk is that those who stay are not the best ones. This is really the challenge to overcome. In this sense, astroinformatics, sorry to say, has no future perspective. If you want to develop in the long term you need to offer permanent positions in that field, having full professors in a position which currently no one has. If we want to be competitive in the field of data driven science we really need to come out with better planning.
This is all. But if you should still be curious, and eager to know more, all the conference talks are available here.
Obviously they are very technical, but they could be interesting to give you an idea of how a reasearch project is developed in our field.