'Many eyes' is a better approach to image recognition than algorithms. When looking at real-universe galaxy merger remnants, and comparing them with simulations, it's better to just have lots of people look at hundreds of possible results (and match them to reality), than to use a machine method. Earlier this week I forwarded details on the Galaxy Zoo 2: Mergers project, which allows you (yes, you! What are you waiting for? Match some galaxies!) do this.
Now, modeling collisional remnants was the topic of my Ph.D. work. I was looking at Graham's Ring, aka AM0644-741, and I ran a swarm of simulations on a Beowulf cluster to attempt to reproduce the system. Since real galaxies take hundreds of millions of years to change, and simulations took perhaps a few hours each, we did have an advantage over reality. But it was a lot of computing time and a lot of time spent with me generating movies from the sims and staring at comparative movies.
The Galaxy Zoo Mergers instigator, who was also my Ph.D. advisor, quipped: "Just a few years ago, a single PhD student could spend his entire dissertation modeling a single interacting system. Oh... you might have heard about this.... :)"
And it's very true. It used to take a supercomputer to do 1 simulation, yet I was able to run dozens-- many dozens. Advances occur, and now we can run them practically in your browser. Moore's Law makes redoing earlier problems (like my collision sim work) easy.
I think the main contribution of my work wasn't the results for AM0644-741 per se, but that I was able to show a smaller sim (of a few thousand elements) was just as good at pinning down a system's origins, as doing the big, high detailed runs of millions+ elements. That we can run small, fast, and frequent, and get more results faster than the big-sim approach.
The main push of Galaxy Zoo Mergers over my work isn't just swarming even more sims, though. It's the web. It's the unfettered interface. It's the ability to get lots of people to compare the results to reality and tell us when there's a match.
The legend of John Henry is that he tried to out-piledrive a steam-powered hammer, succeeding but dying in the attempt. I'm not John Henry... I knew when to let machines take over. I prefer to let people do what they do best-- outthinking the machines.
Alex, the Daytime Astronomer
Tues (for today, in GMT)&Fri here, via RSS feed, and twitter @skyday
Read about my own private space venture in The Satellite Diaries
Galaxy Zoo vs John Henry
By Alex "Sandy" Antunes | November 27th 2009 02:12 PM | Print | E-mail