science

discussion

Team Update: What is a synthetic cluster?

Started by lcjohnso
Default_user
almost 7 years ago

Thank you for your reply.

Besides I would like to know, why a message is displayed when marking synthetic clusters, but when finding one of the 601 clusters, there is however no message displayed (as far as I know) ?

Default_user
almost 7 years ago

We actually had talked about including that as a feature, but frankly ran out of time to implement it before launch!

Default_user
almost 7 years ago

jdalcanton:

We actually had talked about including that as a feature, but frankly ran out of time to implement it before launch!

It would have been fine, but anyway. I will post the suggestion in "Request New Features".

Default_user
almost 7 years ago

To me the synthetic clusters look to me for the most part just like that: synthetic. I'm not keeping score but in the vast majority of cases where I identify a synthetic cluster I know it is synthetic before submitting. They just dont look natural, more like a series of random dots than a cluster which especially in b/w has a kind of background glow of diffuse light that the synthetic ones are missing...

Default_user
almost 7 years ago

I agree with many posters here. Every time I see a really good one, I get all excited about finding something new and it turns out to be one of the synthetic ones. Is the catalog of images so huge that seeing "real" clusters is very low percentage? I think I viewed 50 images over the past few days, and most images had no clusters at all.

Default_user
almost 7 years ago

Some are obvious I agree but I have recently been finding some synthetics that are not so obvious - some very faint and dispersed clusters. My thinking is that only after classifying many images have I become adept at spotting the really tricky ones. I suspect there are even more that I am missing.

Default_user
almost 7 years ago

I agree with Jules. Some days I seem to get lots of synthetics (often groups or in consecutive images) and some days there are very few. I do think that practice helps to identify the small clusters, but feel that it would be useful to be shown the synthetic ones that I've missed, so that I can learn from the mistakes.

It feels a bit like there's a tension between the wisdom of crowds philosophy (restricting input and training) and the science (correctly identifying as many genuine clusters and galaxies etc. as possible to successfully expand the catalogue). Finding the right balance must be tricky... there will undoubtedly be a difference of opinion about the solutions? I agree with zutopian; I too think that it would be interesting to be told if a genuine cluster is known or not. A useful tool to have would be a list of the 600 or so known objects that we could manually check and link to the images -along with a method of doing so!

Default_user
almost 7 years ago

I have noticed something regarding synthetic clusters, how do you account for double classifications of galaxies or clusters in the images when each synthetic image has a duplicate out there? for instance this image http://www.andromedaproject.org.s3.amazonaws.com/subjects/standard/color/B21-F16_28.jpg has two galaxies which will be classified now a synthetic image exists and in that the galaxies will also be classified but they are identical for instance http://www.andromedaproject.org.s3.amazonaws.com/subjects/standard/color/B21-F16_28_sc.jpg

but the galaxies will most likely be classified in both images presented to the same person, I noticed this when check collections and it is noticible when they are side by side, how do you get rid of the double classification from the data? I have not checked but is there a duplicate for each and every image containing a synthetic cluster or are they limited?

small grayish cluster 1st quadrant

not in this identical image less the synthetic cluster

Default_user
over 5 years ago

I have to say I love this project and really appreciate being able to be part of it, Although it's a pity to see so many good suggestions in here that are a year old and still none of the improvements have been implemented.

Please Log In to make comments.