Galaxy Zoo: Mergers – A personal perspective

Now that the launch of Galaxy Zoo: Understanding Cosmic Mergers has been completed, I wanted to give a personal perspective on this project.

For me, this project started twenty years ago  when I was in graduate school.    In my dissertation work, I  modeling the tidal features of interacting galaxies.  I wrote a Fortran code for doing some of this modeling work.   You would set up a run, and then wait hours to see the result.   If it didn’t match, you had to wait hours for the next attempt.

The worst part about the modeling process was getting the “final” result.   Even if you got a close match, you never knew if you had actually found the best match.  It was always possible that a completely different set of parameters was the real solution, and you had just made a mistake.    Even with good fits, you couldn’t tell if you really had arrived at the ‘right’ solution.

Our understanding of galaxy collisions has been limited by the lack of dynamical models.  For example, we know that some galaxy collisions have very high star formation rates.   We also know that almost all extreme star burst galaxies (Ultra-luminous infrared galaxies) have undergone some type of collision.   Why isn’t this sort of reaction the inevitable result of a merger?   It seemed like the answer was always out of reach – unless we can understand the dynamics of lots of collisions.

The java applet developed for Mergers by Anthony Holincheck is the direct descendent of the old Fortran code.  Now you can run the same kind of simulations I ran for my dissertation in fractions of a second.    When Anthony and I first resurrected this code, we immediately tried using a Genetic Algorithm help us converge on the final solution.   It didn’t work.   We couldn’t reliably teach the computers how to recognize a good match.   We could run a few hundred thousand simulations per day, but we never knew if we got the right results.

The idea of using volunteers to help us happened a few years ago.   It was crazy and impractical to imagine volunteers helping out with a project like this.   Even so, a group of us proposed to do.  Of course, our proposal got shot down.  After all, there was no way that this type of thing would work.  How would you recruit such volunteers?

About a year later, I started talking with some of the team from Galaxy Zoo.    You – the volunteers of Galaxy Zoo-  have made the impossible possible.   With your help, we can create the models we need to understand the histories of hundreds of galaxy collisions.   These models will be more reliable than any a single scientist could create.    This result alone would incredibly important.   However, by carefully analyzing your inputs, we eventually hope to train the computers to do thousands of more models.   This kind of man/machine partnership is being planned for a number of future data projects, where computers need help learning how to be scientists.     We will never discover future Voorwerpen or new Peas without your help.   However, in return, we will also never make you do busy work that a machine can do.

Your efforts on the Zoo projects have created a new the way to do science.     This is nothing less than a transformation in how we look at data, analysis, and computing.

Of course, that’s just a personal perspective.

I just posted a new target for you to try.  We are going to be doing updates at 1600 GMT everyday. even Thanksgiving.  Of course, we will keep the old targets live for a week so you can go and revisit them.   This one is a repeat from some of the beta tests.   Getting a perfect model is hard, but getting close is easy.  We didn’t want to make things too difficult – at least for now.   Be assured, we will be kicking up a notch over the next days and weeks.

-John