Space Warps Talk

Info about your data

  • ceperman by ceperman

    I'm interested to know:

    How may images are there in total? I'm wondering if I'm ever likely to get through them all 😃
    Will I see the same image more than once?
    What occurrence of real lenses are you expecting e.g. as a ratio of images to lenses

    Posted

  • JasonJason by JasonJason

    If I understand correctly these (the 127 ) were found in the survey here http://arxiv.org/pdf/1109.1821.pdf and can be viewed here http://kicp.uchicago.edu/~anupreeta/sarcs_sample , I think the total images is somewhere around 400, 000. I am sure someone will correct me if I am wrong. I have also seen images twice but having different reference codes usually when an simulated arc has been added.

    Posted

  • Tom_Collett by Tom_Collett scientist

    @JasonJason is correct.

    If ten people look at an image and almost all say there is no lens, that image is discarded, so you'll never see all 400,000 but I'd be very surprised if you get through even close to that many. It's hoped that future surveys will add more data to spacewarps - I'll be surprised if you can look at 100,000 images before new data is added. Even at 1 second per image you'd need a solid 28 hours of spacewarping to get through 100,000! and it probably needs more than a second to look at an image. (it takes me over a second and I work on lensing everyday - I suspect 2 to 10 seconds per image is reasonable to rule out lensing and ~a minute for anything you think might contain a lens)

    How many lenses are we expecting to see? That's a very good question, but it's surprisingly hard to predict. You need to know about how common and massive potential lens galaxies are and their distance from us, and you need to know the brightness distribution of potential sources and their distances from us, then you need to know what the spacewarps detection limits are. Basically it's tough and I can't give you a definitive answer without a few days of work. @anupreeta might have already crunched the numbers to get an estimate, but I can't do it in my head.

    I can however make an informed guess: I'd expect spacewarps to find all of the 127 already known lenses (plus I think the 127 is only the group scale lenses - the slightly smaller galaxy scale lenses are likely to be similar in number), and probably a similar number again of all new spacewarps lenses. So maybe 0.1% of images could contain a lens - that being said, there will be plenty of maybes, so don't worry if you are marking about 1% of non-simulated images as lenses. If you are finding as many 'real' lenses as simulated, then you are being way too optimistic.

    Hope that's of some help.

    Posted

  • ceperman by ceperman

    It is, thanks.

    I have discovered from a different question I posted that an image is only shown to 10 people, so I guess the number of images available depends to some extent on how active people are. You will never see an image that 10 people have already seen.

    It's also interesting to know a previous analysis has already found 127 lenses.

    Posted

  • robert_gagliano by robert_gagliano

    I presume we are looking at the same CFHTLS T0006 release data. SLS2 found 127 lenses in 400,000 images, that's ~ 1 lens every 3000 non simulated image (and that's the low lying fruit already picked off by computer algorithm). We will detect most or all of the same previously discovered lenses and possibly other previously reported lenses from other lens surveys ect. in the same CFHTLS fields. We also should consider the groups % lens miss rates, positive and negative (not marking one that's there and marking one that's not there). The likely hood of finding new lenses seems very very small. Since 10 people classify each image, the likely hood of being the first to find a new previously unreported lens seems vanishingly small...akin to hitting the Powerball lottery.

    Posted

  • Tom_Collett by Tom_Collett scientist in response to robert gagliano's comment.

    It's not all about being the first person to find something - it's about being part of the team that discovers new things about the universe. Plus you get to learn new cool stuff about how the universe works, that you probably wouldn't think of without 'playing' spacewarps (It's fun right!). And even if you only care about being first the situation isn't too bad - there's probably ~100 unknown lenses in the CFHTLS fields to find and apparently we have had 6744 classifiers to date - so the chances of being first aren't too bad.

    But I strongly encourage you to think of it more as being a part of scientific progress rather than the winner of a race.

    By the way CFHTLS is so good compared to previous surveys that you can almost ignore the possibility of
    lenses in the field being known before CFHTLS - it'll be a handful at most. You're right that the known lenses are the easy ones picked off by computer algorithm, but the algorithms are pretty bad at the moment when compared to the human brain. That's why we need you guys!

    Posted

  • JasonJason by JasonJason

    They wouldn't be using us if they did not think we can improve the computer algorithms, at which point we won't be needed anymore. But I think there will always be some limiting factor that the algorithms can't surpass and brains will win out. Even if it is brains used from the standpoint of "wisdom of the crowds"

    Posted

  • ceperman by ceperman

    If I see a really good candidate, I've been looking to see if it's in the SLS2 list that I was referred to. So far I've seen 3 and commented on the image. Any reason not to do this? I find it useful to see what "real" lenses look like.

    One question I've been wondering about for some while: if a sim appears in an image would it be a guarantee that no real (identified) lens was also in it. It depends on how sims are generated - I thought they would be created by manually manipulating the image, and I assumed one with a real lens would not have been chosen. That's obviously not correct, since image ASW0001e2j contains two sims but also SLS2 candidate SA6. I guess this answers my question although I still don't know how sims are actually created.

    Posted

  • Tom_Collett by Tom_Collett scientist

    No reason not to find out if your candidate is known or not.

    The sims are added by a robot - and have NOT been seen by a human before. There is actually slightly more chance of a real lens being in a frame with a sim than in a random frame. (The robot tries to put sims around likely lenses, and elliptical galaxies cluster in the sky). I think phil and anupreeta are writing a blog post on how the sims are made, so I wont steal their thunder 😃

    Posted