Higgs Hunters Talk

Lots of calibration objects

  • de.hawkeye by de.hawkeye

    Hi,

    seeing a lot of test objects lately. Any specific reason?

    KR
    Michael

    Posted

  • JeanTate by JeanTate

    It's explained in this thread.

    In fact, almost all the objects you classified before were also sims/calibrations ... you simply weren't told.

    Posted

  • de.hawkeye by de.hawkeye

    OK, thanks.

    Posted

  • DZM by DZM admin

    Hey, as we're confirming elsewhere, we have a bug with non-calibration objects being identified as calibration objects. It's true that some objects before were for calibration, but not "almost all" of them... just the normal amount.

    When we went to update the system so that it informed people if they'd seen simulated data, somehow it ended up tagging nearly everything as simulated data. We're working on getting this fixed, and we're sorry for the technical difficulties!

    [EDIT: As per @ttfnrob's post below, this unfortunately wasn't correct. It was based off of what one of the scientists posted earlier, but it would appear that new information has come to light since then. Please see Rob's post below.]

    Posted

  • JeanTate by JeanTate in response to DZM's comment.

    Out of curiousity, how long was Higgs Hunters running, before any sims were 'marked' in such a way that the zooite who'd just classified them would know (either way, sim or not)?

    Posted

  • DZM by DZM admin in response to JeanTate's comment.

    It looks like the code for marking sims was put in on Saturday morning.

    Posted

  • ttfnrob by ttfnrob admin

    Just wanted to try and explain the sims situation. When we launched on Wednesday I thought we had just 1-2% of objects that were simulations. The idea had been to have people blindly classify them and use those clicks to measure the unbiased rate of discovery. This was to be explained in a blog post.

    However:

    The science team explained to me on Friday that, in fact, several the large chunks of data I thought were real data were also simulations. This mix-up was caused a miscommunication internally. So we figured the best way to remedy it would be to just flag when you saw a simulation and add a note to the science pages.

    Right now the Higgs Hunters dataset contains only ~20% real data - the rest is sims. So your 9/10 sims rate isn't too far off, sadly. This is not good and was not the plan. The good thing is that thanks to your comments we've noticed this problem and we can fix it. We'll do our best to get that done in next 24-48 hours.

    One big point to add though: classifying sims is extremely useful too - and the rate of sims will decline as the project goes along. By classifying the sims we're understanding what data Higgs Hunters - as a system - is able to tell us more about. Higgs Hunters is a bit different to some other Zooniverse projects in that there are multiple energies and multiples type of particles that can be seen, so the range of required calibration results is very large compared to Space Warps, for example. The sims show us what can be done, and what cannot be done, with your help on Higgs Hunters.

    My aim is to get the sim rate way down and have it drop over time so that eventually we can run Higgs Hunters on just a few percent sims - as originally planned.

    I'm really sorry for the mix-up, it wasn't our intention to deceive anyone and we hope to have it sorted as soon as possible. I'll post more info when I have it.

    Posted

  • jokergirl by jokergirl moderator

    So wait, which one is correct now? The post by DZM 14 hours ago about that most data was actually real data, or this one (3 hrs ago at point of writing - so fresher), that most data is simulation data?

    Please clarify, as it's really demoralising to be told that we are not working on real data after all.

    Posted

  • ttfnrob by ttfnrob admin

    This, more recent post is correct I'm afraid. However, we've been working on this all morning and I think we will have a much better ratio of sims for you very soon. Updates to follow.

    Posted

  • DZM by DZM admin

    Thanks for clarifying, @ttfnrob. What I wrote before was based off of what one of the scientists had posted yesterday (in this thread), but it looks like we've learned more since then.

    Really sorry to all the volunteers about this. We should have an update very soon.

    Posted

  • koranzite by koranzite scientist

    @DZM Sorry about that, with the flurry of internal emails it was easy to get confused as to whether the sim problem was artificial or not. Rob is quite correct and I miss-communicated the situation, sorry!

    Posted

  • JeanTate by JeanTate in response to ttfnrob's comment.

    Thanks for the clarification ttfnrob.

    If I may, I'd like to repeat what I said earlier/elsewhere, and add something: I strongly urge you/Science Team/etc to write a clear explanation of sims (what they are, why they're used, what they look like, etc). And to put this into the intro material, in a blog post, tutorials, stickied/FAQ thread, etc. And do so ASAP.

    Zooites are, per experience from many Zooniverse projects, more than happy to classify sims ... provided they know what's happening, and why. Treating zooites like some sort of psychology lab rats - by being less than fully honest and open, for example - will lead to unfortunate consequences.

    Too, keep in mind what the demographic profile of zooites is: on average, older and better educated than the US internet population, with degrees biased towards STEM, and a very strong desire to contribute to science. Perhaps, for those with post-grad science degrees and/or present or past STEM careers, explanations etc could go beyond the usual dumbed-down stuff that's all too often served up by the PR departments of universities etc here in the US.

    In short, treat zooites like collaborators, not robots.

    Posted

  • JeanTate by JeanTate in response to DZM's comment.

    Thanks.

    Here's another suggestion: do not introduce major changes over a weekend, especially a holiday weekend (as it was in the US). Or, if you absolutely must do major changes then, at least make sure you have the key players and Dev Team members on hand to swiftly respond if the change doesn't go 100% according to plan (as it almost always will not).

    Posted

  • ttfnrob by ttfnrob admin in response to JeanTate's comment.

    @JeanTate we often put up updates over the weekend because they aren't busy and 99% of the time everything is just fine.

    Posted

  • DZM by DZM admin

    We've made some major fixes and posted a comprehensive update and explanation here.

    Posted

  • ttfnrob by ttfnrob admin in response to JeanTate's comment.

    @JeanTate sims are now mentioned in the tutorial, here on Talk and (as of a few moment ago) on the blog.

    Posted

  • JeanTate by JeanTate in response to ttfnrob's comment.

    According to the Daily Zooniverse's Saturday Status – Higgs Hunters, the traffic on Saturday was ~the same as on Launch Day (Wednesday). I guess Day Four, in most Zooniverse projects, will be busy, whether it's a Saturday or not! 😛

    Posted

  • JeanTate by JeanTate in response to ttfnrob's comment.

    Cool! 😃

    Posted

  • Ptd by Ptd

    Hi DZM
    I admit to knowing practically nothing about particle accelerators or the data they produce, but, please can I re paste something I asked earlier:
    "Also please could the Science Team say in each case what is being simulated, for example particle of type A breaking down into particles B, C & D of which the detector could actually see only C & D, at location E? I recognize that some of us (for example me) aren't really going to understand what those particles are just from the name, but it would be interesting to know, definitively, what some of what we are looking at is."

    And my reasons for asking this are that I'd like to know a little more about how a simulated image is constructed. Does the ST set off a simulated collision in a computer either of a known particle or something they have dreamed up(which would be just fine with me) and then calculate what the detector would see as a result and give that to us raw and without any further manipulation? Or do they hand draw lines in a drawing package, I am guessing something much closer to the former than the latter. But if the former could we not very easily be told what was simulated? For example in PHs we get told the simulated planet's period and size

    At the moment we seem to be being asked to just look for simple ocv's, which in normal view at least, especially if there is more than two lines, are quite easy to spot, if you have perhaps 7+ lines it can be a bit harder to spot if there is just one ocv or two or three almost on top of each other(being able to peal the lines away one at a time with the click of a mouse to see under the mess would sometimes be useful). Its the ocvs with just two lines which are challenging, as that's far more likely to happen randomly with the additional "random looking" white lines that are often in the images. And I'd like to know if perhaps its in those random looking individual lines that some of the gems this project is looking for are to be found.

    MT

    Ptd.

    Please don't think I'm negative about this project, sims or not I'm still clicking away and enjoying it 😃

    Update: I think what I'm really try to get at is I'd like to know if it's possible to learn to spot hidden/in-obvious ocvs by reading between the lines.

    Posted

  • DZM by DZM admin

    Hi @Ptd, I'm not a member of the science team, just the Zooniverse's guy on Talk. 😃

    This is the right board for those questions, though! One of the scientists should be along here before too long, and hopefully they can get you the answers you're looking for.

    Thanks for Higgs hunting; really glad you're enjoying the project!

    Posted

  • andy.haas by andy.haas scientist in response to Ptd's comment.

    What we have simulated is a variety of processes like Z boson + Higgs boson, where the Z decays to two muons, followed by Higgs decay to two long-lived neutral particles, followed by the decay of each of those two neutral long-lived particles to either a pair of tau leptons or bottom quarks. We varied the lifetime (times the speed of light) of the neutral long-lived particles from 1mm to 100mm, and their mass can be 20 or 50 GeV. Since we don't know what new physics is out there, we have to simulate a large range of possibilities (which is partly why we ended up with so much simulated data compared to data. (This is usually real-life in a particle physics analysis at the LHC! 😦()
    The simulations are quite sophisticated and accurate. The particle interactions are calculated according to the rules of the SM + whatever new physics / particles are added. Once these decay to SM particles, they are propagated through a computer-model of the ATLAS detector (kind of like a virtual world in a 3d computer game), where their interactions with the detector material is simulated. The energy deposits in the detector are then read-out by a simulation of the detector electronics. Pileup proton interactions (all the other dots along the center in the slice view) are added, along with noise. Finally the whole thing is reconstructed as if it were real data (almost). It takes about 10 minutes for a modern computer to simulate each of these events!
    For more info on our simulations, see these excellent blog posts from our ATLAS simulation co-leader:

    http://atlas.ch/blog/?p=2353

    http://atlas.ch/blog/?p=2371

    http://atlas.ch/blog/?p=2395

    Posted