Radio Galaxy Zoo Talk

Repeat images (again)?

  • KWillett by KWillett scientist, admin, translator

    A discussion started a few months ago as to whether users were getting any repeat images in RGZ. This definitely shouldn't happen, and isn't in the project design - users are given a randomly-selected image. and our backend software makes sure it's not one that they've seen before. The chances of that are really small when there are tens of thousands of galaxies active, but some of our most active users do enough (and have good memories) that it occasionally happens.

    We pushed a fix to our software to correct for the small number of repeats that users mentioned (see this thread: http://radiotalk.galaxyzoo.org/#/boards/BRG0000001/discussions/DRG000045f) , and that does indeed seem to have fixed the long-term problem. However, there are a couple of more recent sets of classifications I've seen in the data that I've been analyzing, where a user will submit duplicate classifications only a few seconds or minutes apart.

    This is likely a small bug (and one which we can address in the data processing), but it'd be helpful for us to know what your experiences are. Have any of you, since this April, seen the same image within a period of a few minutes or less on the site? If so, please let me know in the comment thread. This will help our developers determine whether it's a bug in the way that the images are being delivered to you, or in the way that they're being recorded in our database. Thanks!!

    Posted

  • KWillett by KWillett scientist, admin, translator

    Since this is a rare bug, btw, it likely occurs most for those users who have done lots and lots of classifications. @antikodon, @WizardHowl, @JeanTate, @Dolorous Edd - anything you've seen recently?

    Posted

  • antikodon by antikodon in response to KWillett's comment.

    Yes, I have come across repeats image. I think that the problem with my ISP (cmd ping 54.231.2.196 /176.32.101.4 sent=4 received= 0 lost= 4)

    Posted

  • KWillett by KWillett scientist, admin, translator

    Thanks, @antikodon. Can you be more specific about the experience? For example, how much time was there between images, and did you classify any other objects between seeing the duplicates?

    Posted

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

    Thanks Kyle.

    I don't know about the "have done lots and lots of classifications" (my count is surely quite low!), but I have seen no 'short cycle' repeats since April. I have been especially keen to make sure I record any such; I'll certainly update this thread if I come across any.

    Posted

  • JeanTate by JeanTate

    Not sure if this counts, but xDocR* (formerly DocR? the extended version of DocR?) reports getting 'the tutorial image' ARG0003r15 twice (June 7 2014 6:47 AM): "got the tutorial source for a second time."

    *written in code because Talk turns this into xDocR ! 😮

    Posted

  • KWillett by KWillett scientist, admin, translator

    Second version of @DocR - we're doing some tests on the site for calibration under a different name. Thanks, Jean.

    Posted

  • JeanTate by JeanTate

    antikodon reports getting ARG00026zh to classify a second time; i.e. a repeat. And it's of the 'long cycle' kind, that the April fix was supposed to eliminate.

    First timestamp: June 3 2014 2:13 AM

    Second: June 9 2014 11:37 PM

    Posted