science

discussion

Team Update: What is a synthetic cluster?

Started by lcjohnso
Default_user
almost 5 years ago

From today's blog entry:

In addition to real clusters that appear in the images, the science team has inserted highly-realistic synthetic clusters into a portion of the images. These objects are essential to the scientific results we hope to obtain, as they allow us to determine which kinds of clusters can be detected in the images and which will be overlooked. Careful measurement of these limits will allow us to derive how Andromeda’s stellar clusters have evolved over time. Having you find these clusters is just as important for many of our science goals as the real clusters!

This task requires a significant number of synthetic test objects to make sure we can model how our ability to identify clusters changes due to different competing effects. These factors include cluster size, cluster age, and a cluster’s location in Andromeda and the density of stars in the image.

These synthetic clusters are also a way to provide some feedback on how you’re doing with cluster identification. Unlike the real clusters, we know where these objects are located in the image ahead of time. That allows us to tell you when you’ve correctly identified a synthetic cluster. If you’re finding the synthetic clusters, you are likely identifying new, real clusters as well!

Default_user
almost 5 years ago

How many synthetic clusters are likely to be found? Is there any way to measure how "good" one is performing? so frustrating not to know how many of the clusters you identify are real ones... :(

Default_user
almost 5 years ago

I would like to thank the team for allowing us the opportunity to participate. As a long time amateur astronomer one of my aspirations was to be able to pick out clusters in this galaxy through a telescope.

Not to be insulting but to me a fair number of the "highly-realistic synthetic clusters" look synthetic. So much so that at first I was not identifying them as clusters. Also I quickly became weary of identifying image defects and stopped doing so early on. Sorry about that.

How many images are there and is there a way to find out how many we have viewed? As the others have mentioned there seem to be a lot of repeat images.

Default_user
almost 5 years ago

lcjohnso:

From today's blog entry: ...

These synthetic clusters are also a way to provide some feedback on how you’re doing with cluster identification. Unlike the real clusters, we know where these objects are located in the image ahead of time. That allows us to tell you when you’ve correctly identified a synthetic cluster. If you’re finding the synthetic clusters, you are likely identifying new, real clusters as well!

Well, I think, that there is a better way to check, how good we are in identifying clusters.:

As far as I know, there are 600 clusters, which were identified by astronomers before. I guess, that those images are not displayed for classification. If so, I think, that those images, which show previously indentified clusters, should also be displayed for classification.

I know something like that from planethunters.: As far as I know, some of the displayed light curves show transits of planet candidates, which were discovered by the Keppler Team before, but besides they displayed some simulated light curves.

Default_user
almost 5 years ago

@zutopian: I don't know, of course, but I think it would be silly to deliberately exclude the ~600 known clusters from the set of AP images served up to us zooites. And I very much doubt that such exclusion has been done. Why? Because some quite prominent globular clusters have turned up, they were all surely known well before the AP got started (big globulars are pretty easy to find).

What's more likely is that, among the ~600, are several (dozens? hundreds??) which are not in the (to date) AP footprint, or even the PHAT footprint.

Default_user
almost 5 years ago
zutopian in response to JeanTate

JeanTate:

@zutopian: I don't know, of course, but I think it would be silly to deliberately exclude the ~600 known clusters from the set of AP images served up to us zooites. And I very much doubt that such exclusion has been done. Why? Because some quite prominent globular clusters have turned up, they were all surely known well before the AP got started (big globulars are pretty easy to find).

What's more likely is that, among the ~600, are several (dozens? hundreds??) which are not in the (to date) AP footprint, or even the PHAT footprint.

In planethunters there was an information given at the end of the classification of Keppler Team candidates.

So if in the Andromeda Project, images of clusters from the paper http://arxiv.org/abs/1204.3091 are indeed displayed for classification, there should be an information given at the end of the classification!

Default_user
almost 5 years ago

It would be nice that when we are done with an image that it would not only tell us that we found a synthetic cluster but show us the ones we missed and show us known clusters that we missed so we can learn to better identify them.

Default_user
almost 5 years ago

@rockspringer just my opinion!!

Default_user
almost 5 years ago
zutopian in response to JeanTate

JeanTate:

@zutopian: I don't know, of course, but I think it would be silly to deliberately exclude the ~600 known clusters from the set of AP images served up to us zooites. And I very much doubt that such exclusion has been done. Why? Because some quite prominent globular clusters have turned up, they were all surely known well before the AP got started (big globulars are pretty easy to find).

What's more likely is that, among the ~600, are several (dozens? hundreds??) which are not in the (to date) AP footprint, or even the PHAT footprint.

Following information is given in the Guide:

Known clusters

Some of the clusters that you identify may have been found previously. There are a number of catalogs of star clusters that have been created from lower resolution ground-based observations. These catalogs include many of the brightest clusters, but miss most of the lower mass clusters that you can identify. We know this because in our initial work on this dataset (Johnson+ 2012), our team found up to four times as many clusters than had been previously cataloged. Most of the clusters you identify will never have been identified before.

So I ask :"How many previously found clusters are displayed?" It is still not clear to me, if the 601 clusters from the initial work on the dataset are also displayed in this project or not?

Asking for a comment by the science team!

Default_user
almost 5 years ago

Hi zutopian -- Thanks for your question, and sorry there's sometimes a delay before we get back to you.

For the Year 1 catalog paper, we searched ~25% of the total dataset and found 601 high-quality clusters. We do include these Year 1 images as part of the image set served up through the Andromeda Project (AP) site for a couple reasons:

+At the conclusion of the AP work, we would like one single catalog of objects that was constructed in a uniform way. This will allow for better scientific results in the future because there won't be any inconsistencies in how we gathered the information as we move from one region of the galaxy to another.

+We expect the results from the AP to be higher quality than the Year 1 catalog due to the increased search redundancy!

+Similar to zutopian's suggestion, we plan to use the comparison of the published Year 1 catalog as a way of checking the quality of the AP catalog. However, as I eluded to in the previous point, we actually expect the AP results to trump the Year 1 catalog in the end, so we will use a combination of both the synthetic cluster results and the comparison to our previous work to perform this training and calibration.

Great question - thanks for the interest!

Please Log In to make comments.