The Finale of Merger Zoo

Working toward Fitness – 

Over the last year, we have been pretty quiet at the Merger Wars site.   However, we have been extremely busy analyzing the data that you have created.     So far, the Merger’s Applet has been used to view over 3 million simulations of interacting galaxies.   We have estimated it actually simulated more than 100 million other systems that weren’t viewed by our users.

Of the 3 million simulations viewed, around 60 thousand were selected by the volunteers as interesting. Initially we thought the Evaluate activity within the applet would be sufficient to help us identify the top simulations for each pair of galaxies. However, with millions of simulations to sort through, across tens of thousands of sessions, we discovered that our initial plan was not sufficient. That’s when we decided to add the Merger Wars activity.

In Merger Wars volunteers judge a series of head-to-head competitions to determine which simulation is a better match to the target image of a galaxy merger. Over time, as the simulation competes multiple times, it earns a win/loss record. The percentage of times the simulation has won its competitions can be treated as a fitness value. A value of 1 is a perfect score, all wins, and a value of 0 is a terrible record of all losses. With over 800 thousand Merger Wars competitions judged, our volunteers were able to help us assign a fitness value to each of the 60 thousand selected simulations. These fitness values allowed us to further refine our models for each merger. In total, we identified 290 top simulations for the combined set of 54 pairs of galaxies. However, we need some final help finding the very best model for each system and finding out which collisions have the very best models.

 One Orbit to Rule Them All-

Ideally, there would be a single set of orbit parameters to describe the paths the two galaxies take when flying past each other, and eventually towards their ultimate merging into one galaxy. It is difficult for researchers to know for sure if they have found the single best set of parameters. Is there a better set of orbit parameters? Are their multiple sets of equally good parameters? Are there no good sets? We can call this problem the issue of determining uniqueness. The volunteers for Merger Zoo have achieved an unprecedented level of study for each of these 54 systems. Typically researchers will look at a few dozen to a few hundred simulations of interacting galaxies and pick the best orbit from that sample. Together, we’ve reviewed on average over 50 thousand simulations for each pair and selected over 1000 simulations for further study. We’ve taken the multiple sets of orbit parameters identified for each system and examined them to see how well they’ve identified a single, best-fit orbit. When we look at the entire sample, we don’t see a single orbit. However, if we begin to exclude some of the sample by filtering out the low fitness simulations, we see the range of orbit parameters becomes smaller. If we increase the fitness value used in that filter, we continue to see smaller and smaller ranges of values. In this manner, we can say that we see convergence towards a small range of values for each of the orbit parameters. Arp 82, the image from the top of the post, is a good example of this convergence.    For each population we show a box plot describing the distribution of the parameter. The box represents the range of data from 0.25 to 0.75 of the population. The horizontal line is the median, and the thin whiskers show the outliers. The populations shown are all states viewed by the users, all states selected by the users, and then several populations filtered by fitness to include the to top 50%, 25%, 10%, 5%, 2%, and finally the top 1% of simulations by fitness. The distribution of values describing the time of closest approach demonstrates some convergence. We see that the applet sampled a range of orbits that had times of closest approach ranging between 60 to 600 million years ago. By the time we filter to just the top 1%, we see the range is now only 100 to 250 million years ago with the likely range of 120 to 162 million years ago. Hancock et al find a time since closest approach of around 200 – 250 million years.  The ratio of masses between the two galaxies converges in the same way.

Your Help is Needed-

The big thing we need help with is figuring out THE very best model of each system and comparing models of different systems.    The Merger Wars site has a couple of new interfaces that are now posted.   They include:

  • Pick the Best: find the best model for each system
  • Simulation Showdown: comparing the simulations from two different galaxies to find out which systems have the best models
  • Merger Wars – HST :  the last batch of Merger Wars results from non-Sloan images
  • My Mergers: This is a new update that shows your contributions to the project.  [Note: you need to be logged in for this link to work.]

After we get this final set of data, we will be archiving the site and writing a set of papers based on this work.  However, your help is really needed with this final part of our analysis.

A Zoo of Mergers-

The image below is a combined image of 54 thumbnails showing all of the SDSS galaxy pairs studied in detailed by our Merger Zoo volunteers. Clicking on the image below will take you to our updated Gallery. From that page you can click on each individual thumbnail to see the top simulation results.    You have done an amazing job with this project.   Thanks so much for your help.

Galaxies Modeled in Merger Zoo

John and Anthony, The Merger Zoo guys


About The Zooniverse

Online citizen science projects. The Zooniverse is doing real science online,.

10 responses to “The Finale of Merger Zoo”

  1. c_cld says :

    Could you correct a typo on SDSS ID 1st col, 3rd row: the object is not
    587738569246376675 but 587738569246441487.

    I wonder why in your sample there are less than a half of known redshifts for the two interacting galaxies: I thought the distance between the two nuclei was an important parameter for the gravity potential simulation for your checking with reality.


    • John says :

      In some cases, we used the redshifts from other sources. A fair number of these systems have additional observations and were cataloged in the Arp Atlas. The galaxies in the pairs within our sample had redshifts that were more than a few hundred km/s different, so they should in the same region of space.

      I will see if I can get the results page corrected in the next few days.

  2. echo-lily-mai says :

    Here is a link to c_cld’s question on the forum

  3. echo-lily-mai says :

    The link in your blog to “My Mergers” under “Your Help is Needed-” is not working

    • John says :

      Thanks! Apparently you can’t get there unless you are logged into the site. The link is on the login page, underneath “Hello USERNAME!”

  4. echo-lily-mai says :

    Thanks for your reply John. I logged in and it says “Our apologies, calculating this value was more difficult than originally thought.”

  5. zutopian says :

    – I did a post, which refers to c_cld’s post, on the forum today.
    – I logged in and clicked “My Mergers”, but it doesn’t work:
    “Website restore error”.(I use Internet-Explorer 9.)
    The classification options work fine!
    – I am astonished, that none of the 54 SDSS GPairs matches with the 12 “GalaxyZoo Binary AGN Candidates”(if I checked right), because Kevin had written in a blog post from 2010 following:
    “Oh and you might notice some of the targets in the Merger Zoo in the near future. We’ll need your help to fully understand them….”

  6. Jean Tate says :

    Wednesday (4 April)’s Object of the Day is about this final Merger Zoo challenge:

  7. mhannigan says :

    I was just thinking about this. I want to play with merging galaxies again. I wish this were still an ongoing project!

Trackbacks / Pingbacks

  1. Models of Merging | Galaxy Zoo - April 26, 2016

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: