Galaxy Zoo on the Naked Astronomy Podcast
The July 2012 of the Naked Astronomy podcast includes an interview I did with them (at the UK National Astronomy meeting this spring) about Galaxy Zoo and why it’s such a great way of learning about galaxies in the universe.
Machine Learning & Supernovae
This post, from Berkley statistician Joey Richards, is one of three marking the end of this phase of the Galaxy Zoo : Supernova project. You can hear from project lead Mark Sullivan here, and from the Zooniverse’s Chris Lintott here.
Thanks to the efforts of the Galaxy Zoo Supernovae community, researchers in the Palomar Transient Factory collaboration have constructed a machine-learned (ML) classifier that can reliably predict, in near real-time, whether each candidate is a real supernova. ML classification operates by employing previously vetted data to teach computer algorithms a statistical model that can accurately and automatically predict the class for each new candidate (i.e., real transient or not) from observed data on that object. The manual vetting of tens of thousands of supernova candidates by the Galaxy Zoo community has provided PTF an invaluable data set which could be used to accurately train such a ML classifier.
The ML approach is appealing for supernova vetting because it allows us to make probabilistic classification statements, in real-time, about the validity of each new candidate. Further, it allows the simultaneous use of many data sources, including both new and reference PTF imaging data, historical PTF light curves, and information from external, on-line sources such as the Sloan Digital Sky Survey and the U.S. Naval Observatory. In total, our automated ML algorithms use 58 metrics about each supernova candidate, all of which are available within seconds after PTF detection of the candidate. These metrics—features in ML parlance—are fed into a sophisticated algorithm, which uses the aggregate of information from more than 25,000 historical supernova candidates which were rated by the Zoo to instantaneously determine whether each newly observed candidate is a supernova.
Human vs Machine scores for Supernova candidates
False Positive rate and missed detection rate for human and machine classifiers.
Our “ML Zoo” has been operating since the beginning of 2012 and has been thoroughly tested against the Human Zoo scores. We found that the ML Zoo scores correlate reasonably well with the average Human Zoo scores for 7000 supernova candidates observed during the first 3 months of 2012 (Figure 1). We also discovered that the ML Zoo is more effective at finding supernovae. In Figure 2 we show a plot of the supernova false positive rate (% of non-supernovae that were classified as supernovae) versus the supernova missed detection rate (% of confirmed supernovae that were classified as a non-supernovae) by both the Human an ML Zoos for 345 spectroscopically confirmed supernovae from 2010. Indeed, the ML Zoo achieves a smaller missed detection rate at each false positive rate.
Joseph Richards works in the Statistics and Astronomy departments at
the University of California, Berkeley as an NSF-sponsored
postdoctoral researcher supported by an interdisciplinary
Cyber-enabled Discovery and Innovation grant. His main area of focus
is astrostatistics and he holds a Ph.D. in Statistics from Carnegie
Mellon University. In his academic research, he has developed
sophisticated statistical and machine learning methodologies to
analyze large collections of astronomical data.
Supernova Project Retires
This post, from project lead Mark Sullivan of Oxford, is one of three marking the end of this phase of the Galaxy Zoo : Supernova project. You can hear from Joey Richards of PTF here, and from the Zooniverse team here.
Since August 2009, Galaxy Zoo Supernovae has been helping astronomers in the Palomar Transient Factory (PTF) find exploding stars, or supernovae, in imaging data taken with a telescope in Southern California. This project has been tremendously successful – Galaxy Zoo Supernovae has uncovered hundreds of supernovae in the PTF data that would otherwise have been missed. These discoveries have directly resulted in scientific publications, with many more in the pipeline, and have been observed on telescopes all over the world, including the 4.2-metre William Herschel Telescope. For example, my colleague Dr. Kate Maguire’s paper includes 8 supernovae found by the Zoo, which were subsequently observed using the Hubble Space Telescope. This allowed her to examine the ultraviolet properties of several thermonuclear ‘type Ia’ supernovae, the same type as those used in the original discovery of dark energy and the accelerating universe. The ultraviolet is a probe of the composition of the exploding star, and allowed her to test whether type Ia supernova properties change with time as the universe ages and becomes enriched with heavy elements.
But – all good things must come to an end. One of the goals of Galaxy Zoo Supernovae was to use the Zoo classifications to improve the algorithms that surveys such as PTF use to find supernovae automatically. And the good news is that, after two years of hard work, we have managed to do just that. The full details are explained in a separate blog posting by Dr. Joey Richards at the University of California at Berkeley.
I’d like to take this opportunity, on behalf of everyone involved with PTF, to thank you all for your time and effort in classifying these supernovae for us. We realise how much effort you’ve put in, and it has been very much appreciated.
For those of you who have become addicted to supernovae, don’t panic – there may be further supernova-related projects in a few months time. In the mean-time, watch this space for more publications based on Galaxy Zoo Supernovae discoveries!
Galaxy Zoo Science Wordl
I’ve given a couple of public talks recently on results on galaxy evolution from Galaxy Zoo (at the Hampshire Astronomical Group, and the Winchester Science Festival) and one of the things I like to point out is the quantity and variety of science results we’re getting out. To illustrate that I made the below wordl of words appearing in the abstracts of all the peer reviewed science papers the Galaxy Zoo science team have put out.
This is based on the 30 papers about astronomical objects submitted up until July 2012. I just missed Brooke’s first financial reform paper submitted by a day or two, and I love that this was out of date just as soon as I made it. 🙂
I realise I should sort out things like per cent – percent, and galaxy, Galaxy, galaxies technically being the same thing. But still I think it’s interesting to look out.
Update: Peas are a Mess!
Hi all,
About two weeks ago, a group of astronomers led by Ricardo Amorin posted a new paper on the peas to astro-ph. They used the giant Gran Telescopio Canarias (GranTeCan or GTC) to take really high-quality spectra of some of the peas. What they find is amazing, but not entirely unexpected. We already knew from Carie Cardamone’s paper that the peas are extremely intense starbursts, that is, they form more stars relative to their mass than any other kind of galaxy in the nearby universe. Now, Amorin et al. show that they are a real mess:
A GTC spectrum of a pea from Amorin et al. (arXiv:1207.0509) zoomed in on the Halpha emission line. The Halpha line comes from gas ionised by the powerful radiation from very young stars. The high-resolution spectrum clearly shows that the single Halpha line is actually due to several different components.
The Halpha emission lines of the peas, once studies at high resolution and signal-to-noise, show that they are actually composed of several different lines. The Halpha line is generated by the powerful ionising radiation from young, massive stars hitting the surrounding gas. The multiple lines mean that the peas have several chunks of gas and stars moving at large velocities relative to each other.
This makes sense from what we know from the few peas that have nice Hubble images.
Hubble image of a pea galaxy (Amorin et al. (arXiv:1207.0509)) showing that it actually consists of multiple components; it’s a real mess!
The multiple Halpha lines are almost certainly from these multiple components and suggest that the gas (and stars) in the peas are effectively a turbulent mess. Some of those clumps whiz past each other at over 500 km/sec. Yes, km/sec. Some of the Halpha lines are also broadened suggesting that really energetic events are occurring inside those star-forming clumps, such as multiple supernova remnants or powerful Wolf-Rayet stars.
You can get the full paper as PDF or other formats here on arxiv.
Something rich and strange – Hubble eyes NGC 5972
We just got the processed Hubble images for NGC 5972. This is a galaxy with active nucleus, large double radio source, and the most extensive ionized gas we turned up in the Voorwerpje project. We knew from ground-based data that the gas is so extensive that some would fall outside the Hubble field (especially in the [O III] emission lines – for technical reasons that filter has a smaller field of view). We expected from those data that it would be spectacular. Now we have it, and the Universe once again didn’t disappoint. Another nucleus with a loop of ionized gas pushing outward (this time lined up with the giant radio source), twisted braids of gas like a 30,000-light-year double helix, and dramatically twisted filaments of dust suggesting that the galaxy still hasn’t settled down from a strong disturbance.
Here’s a combination of the Hα image (red) and [O III] (green) data, with the caution that neither has been corrected for the contribution of starlight yet. The image is about 40 arcseconds across, which translates to 75,000 light-years at the distance of NGC 5972. This gives the team plenty to mull over – for now I’ll just leave you all with this view. (Click to enlarge – you really want to.)
![NGC 5972 from HST in [O III] and H-alpha](https://blog.galaxyzoo.org/wp-content/uploads/2012/07/ngc5972o3ha.jpeg "NGC 5972 from HST in [O III] and H-alpha")
NGC 5972 from HST in [O III] and H-alpha
What we still don’t know
I used to think that science was about discovery, about adding certainty to what we know about the Universe. Discoveries happen, of course, but I’ve learned that the really exciting stuff happens not when we expand our knowledge, but our ignorance; progress is measured in the number of unanswered questions we have. After all, any good result raises more of those than it answers.
I have this in mind because today is the 5th anniversary of the launch of Galaxy Zoo, and it’s tempting to write about how we – with your help – have magnificently fulfilled the vision we had back in 2007. After all, in that first story on the BBC news website; a youthful version of me chirps that “We hope that participants in Galaxy Zoo will not only contribute to science, but have a lot of fun along the way”. Science? Check. Fun? Check..
But did we really understand what we were getting into? Certainly not. We’ve rehearsed before the story that we didn’t understand the size of the response we would get, nor the undimmed enthusiasm for sharing in exploring the Universe that still motivates volunteers today. But on launch, we didn’t realize we needed this blog to explain what we were doing with the clicks, nor the forum; which (thanks to the efforts of Alice Sheppard and team) has played such an important role in defining Galaxy Zoo. We didn’t realize that detailed classifications, of bars and three-armed spirals, of bulgeless disks and merging galaxies, were possible, nor that thanks to the Hubble Space Telescope we’d end up exploring the distant Universe, peering at blue blobby galaxies in a mixture of interest, awe and frustration.
We didn’t realize that spontaneous discovery, serendipitous exploration of the cosmos would come to provide some of the most entertaining and scientifically valuable results from the project. From the Voorwerp, to the recent Hubble images of the Voorwerpjes; (another hit my inbox this morning – watch this space) through to the Peas which are now attracting rather a lot of attention). On a personal note, on that July morning in 2007 I didn’t know most of the people who would lead this scientific return – Kevin, of course, was still recovering from classifying 50,000 galaxies himself, and Kate Land and Anze Slosar provided sterling support, but Steven Bamford and Karen Masters in particular had yet to step forward into their leading roles. Much of this science will be celebrated at a one day meeting at the Royal Astronomical Society next year on ‘Galaxy Morphology in the era of large surveys’ – mark your diaries for May 10th! The most exciting work to be presented at that meeting probably doesn’t exist yet – I suspect we’ll still be puzzling over exactly what bars do to galaxies (or vice versa), and arguing about exactly how black holes grow, but all we have at the moment is an ever-growing pile of questions. Which is, of course, exactly as it should be.
We also didn’t know what we didn’t know when it came to development. The original site worked brilliantly, thanks to the efforts of Phil Murray; and Dan Andreescu, but probably the biggest change over the last few years has been the arrival of Arfon Smith and his merry band of developers. That, of course, has spawned a whole new Zooniverse, which has sent us hunting for supernovae, planets, looking for bubbles and even listening to whales. In that manic expansion, Galaxy Zoo has occasionally been left behind, but I’m pleased to say that a new site is on the way. By the middle of August, a brand new site will be serving up images of new galaxies, both from the deep CANDELS survey and, returning to our roots, from the latest data release of the Sloan Digital Sky Survey. We do need to have a few more clicks on the existing site, though, so anyone who classifies in the next two weeks on Galaxy Zoo will be rewarded with early access to the new site (whether or not you’re still reading at this point).
So much for the next few weeks. What of the next few years? I could tell you that as Galaxy Zoo has established citizen science as a standard way of doing astronomy, you’ll see many more projects from us exploring pretty much every aspect of the Universe. I could tell you that I suspect that live interaction with data fresh from the telescope; is going to be increasingly important as the amount of data available to astronomers reaches at least 120 terabytes by the end of the decade. I could spend hundreds or thousands of words convincing you that advanced tools are key, that we’re going to need many more people to follow the lead of the denziens of the forum and get deeply involved in the science that lies beyond clicking. And I could tell you of our determination to finally crack a means of getting Galaxy Zoo firmly into the classroom, but the truth is anything could happen. And that’s just the way we like it.
Chris
P.S. To anyone who has taken part in the last 5 years – thanks a million. Now go and get classifying.
Galaxy Zoo at the European Week of Astronomy and Space Science
The Galaxy Zoo science team is well represented this week at the annual European Week of Astronomy and Space Science, hosted this year at the Pope’s University (or more properly Pontifica Universita Laternase) in Rome, Italy.
It is a beautiful location for a conference
with the most amazingly decorated lecture theatre I’ve ever been in
and just up the road from the Colluseum
A session on the first day on the Structure of Galactic Discs perhaps explains the interest of many of us on the Galaxy Zoo team. I spoke in that session on my recent results looking at bars and the atomic gas content of nearby galaxies.
Brooke Simmons (now settling in as a new postdoc at Oxford after finishing her PhD at Yale recently) had a poster on some work I’m sure you’ll hear about soon about some very interesting totally bulge free disc galaxies which still have actively growing supermassive black holes in their centres.
And Portsmouth PhD student, Tom Melvin (who is working with me) had a poster on his work using Galaxy Zoo: Hubble data to look at the redshift evolution of the bar fraction (more on that very soon too I hope).
Finally, talking in the session on interacting galaxies which runs tomorrow will be Kevin Casteels from Barcelona (who we all must congratulate on his very recent PhD) who has been working mostly with Steven Bamford on morphological signatures of closely interacting pairs of galaxies (arxiv link to paper, a blog post has been promised).
We all had a lovely (and typically late Italian) dinner together on Monday night – along with a Galaxy Zoo baby: Alia (Kevin’s daughter).
Designing the Lens Zoo: Have Your Say!
Over at the brand new Lens Zoo project blog we are starting to document our progress towards a new zoo being built this autumn – a zoo for finding gravitational lenses! A small but dedicated band of lens hunters has been active on the Galaxy Zoo forum for several years – while we are still working with them on the objects they have found so far, we’d also like to help them extend their science investigations by providing new data and new tools in the new zoo. If you’ve spent time on the forums looking for rare objects like lenses, or have stumbled across something unusual that has led to an interesting adventure in the Galaxy Zoo, we’d like to hear from you – we’re trying to figure out how to help you make one-in-a-thousand discoveries.
So: what features should the Lens Zoo website have, to help us find as many lenses as possible? We are planning a workshop in mid-July to discuss the interface and tools for the new Zoo, and to give us something to talk about, we’d love to hear from all you lens-hunters out there. We’ve setup a web form for you to send us any ideas about functionality or tools that you think would be useful in finding lenses. Here’s the address in full:
We’ll go through all your ideas when we meet up in Zurich, and keep you posted on the Lens Zoo blog!
Stay tuned, and thanks for your help.
Phil, Aprajita & the Lens Zoo team
Merger Zoo
Merger Zoo has come to a close.
Since the project started, we have had over 27,000 volunteers contributed their time supporting this project. Volunteers have reviewed over 3 million simulations. Out of this huge number of new simulations, we have been able to find the best models for each of 60 different merging galaxy systems using the data you generated in the Merger Wars and Simulation Showdown interfaces.
The two images above show an overlay for two of the best simulation from Merger Zoo. As the image fades between the simulation and the astronomical image, you can see how closely we matched the shapes of the real Merging Galaxies. Of course, the underlying purpose of this Merger Zoo was not to make pretty models. We are now in the process of analyzing the incredibly rich data set that has been generated to address a number of scientific questions.
The first paper we are working on addresses a simple question – how well can the orbit of the mergers be constrained from the shape of the tidal features? For decades we have been assuming that there is a true “best fit” orbital match for interacting galaxies. However, Merger Zoo has directly put this to a test. As an example, take a look at the plot below. The red line shows the distribution of different disk crossing angles (inclinations) from all the orbits that were viewed by our volunteers. The green line shows the states that were actually selected and survived the first rounds of the Merger War’s competition on the site. Even though none of the volunteers ever look at the inclination angle, the states our volunteers have selected are converging toward a single best angle.
The uniqueness of merger orbits is only the first of many of papers that we are working on. We are also looking how the star formation rates in mergers depend on the orbits between the two galaxies. We have come a long way on this analysis, and seem to be close to some nice results. We are also looking at ways to automatically model merging galaxies using computer vision. The Citizen Science data from Merger Zoo will be used as the training set for the computer vision program.
When Anthony and I look at this Merger Zoo today, we are thrilled with the quality and quantity of the data that you have generated in this project. I have wanted to have models for a large system of galaxy interactions for decades to test some of these difficult questions. Without your help, creating this set of models would not have been possible. With all these data that has generated, the hard work for Anthony and I is really just beginning. We will be spending our time to make sure we turn your time and effort into scientific knowledge. Of course, we will keep you informed as this process continues and results are published.
Thank you for all your help in this project!
John and Anthony, The Merger Zoo Team
Galaxy Zoo 