Merger Feedback. Post your comments here.

[Zeus2007]

I've only done a couple, but I find the enhance option not confusing, just a little hard.  I guess that's all about getting used to it.  Anyway, do we need to use both the explorer and enhance methods to prepare our work before sending it in? As I understand it, the enhance window is there to help our best sims to get as close as possible to the target merger, so it's an important tool in the project.  In my case, it can take me over half an hour to get everything ready and I find the cross thingy hard to move.  It usually says up to increase and right to decrease...I don't know if my mouse just doesn't it or they haven't known each other long enough to cooperate.  Other than that, the rotating option is awesome I spend a bit of time playing around with it just to get the full picture.  Of course, then I get fustrated when I can't get it back to the way it was before!

[graham d]

Hello Zeus- it sounds that you are experiencing some of the same operations we all are. For what it's worth, and it's my own approach, here's my approach that is still evolving.

You have clicked on several or even several hundred simulations and then click on one simulation to enhance. This simulation appears in the right hand box at the top.
 
You may then click on whatever options appear in the left hand box. Once you move that often very sensitive cursor you may see a better fit. My advice is SAVE. It appears in the LH box below the original simulation. Right click on the saved simulation since you might care to make it the best fit, or make it a favourite etc. Then move the cursor again. If you get a better fit then SAVE. I tend to think you cannot save enough. Why? Often when you move the cursor back to its original position it doesn't give the former pattern!

Once you rotate the image on the screen my advice is far more often than not you will unable to recover the original orientation. Hence, if you have spent tens of minutes twiddling before you rotate on the screen and haven't SAVED you feel a dumb idiot. Obviously the right box gets cluttered with far more than eight saved simulations. Keep checking those saved. You might want to prioritise.

Once you click on depth- let me say SAVE before you click on this option. Why? Because you can never get back to the original pattern. Often the merest shift of the cursor takes you into chaotic patterns. If you have SAVED all is happy, you can recover. It's great to select Depth and then orientate on the screen to see the 3D effect, and how  the depth cursor affects the trajectory's approach to the main galaxy. Rarely can you get back to the original pattern. Hence- the SAVE function is vital. If it's starting to look like a much better fit you have it SAVED and can always right click to prioritise as a best fit when it goes to the top of the list or make it a favourite. Don't forget that original simulation which may be getting lost down the list. Why?

Suddenly- EVAluate appears. Drat!- I forgot to save it- just a few more tweaks and I could have made it better. DON'T attempt to click back to the previos page on the browser.

Three best fits- well make them although the selection doesn't look as good as your last fit, which of course you forgot to SAVE. Then select that single best fit. You have reached Level1. Select whatever option, size or mass are best and move the cursor ever so slightly and SAVE. Why? Say you select depth or angle and move the cursor ever so slightly. Sometimes you are in an area surrounded by chaos and never can recover that delightful fit. Remember you cannot SAVE enough- it's your lifeline- it's like striking in a piton and attaching the carabiner on your quickdraw or inserting your friendly cam on the route upwards to phase space heaven.

Eventually you get to level 2. I'm still not sure if this is a top of the pitch with more to come as your skill improves. You can still fall off though but you are protected by that last piton; hopefully it holds. Tournment completed. Maybe it took an hour or longer. Remember to hit the red stop button and follow the instructions. Don't just exit from the browser. Perhaps you want to continue? By this time your best fits have been uploaded and you rememberd to commit at least one of those best fits to Favourites. There are several peaks out there, some narrow and sharp where the cursor appears like a pair of hob nailed boot, others broad and undulating perhaps where the cursor is like a a tight pair of skin fitting sticky rubber climbing shoes which would better suit those narow sharp peaks. So don't blame the tools, they will better evolve with time.   
 

[graham d]

That's life. Already signed in I returned to the mergers site and clicked on MY MERGERS. Must have been a snow storm as all my favourites are gone! That's life. I'm smiling though. Although I'm no longer able to climb those old peaks, as my tools are worn out, I'll abseil down these new peaks and start again- this time doing it better. There are many different routes to the top of these mountains so experiment your own way; explore, select, enhance, SAVE, enhance SAVE- every few minutes strike in that piton.

[jules]

 
I agree entirely Graham. SAVE before each twiddle! The number of times I have got one looking almost perfect to think "I wonder what it would look like if I flipped the secondary....?" then clicked the button having forgotten to SAVE!!!

Getting used to the idiosynchrasies is all part of the game.

Oh - and My Mergers are still intact and all favourites still in place. (Though not all resemble the mergers I actually saved.) And I can confirm there is a level 3.

[MKS]

First, understand that my computer knowledge is limited.
I was working on the merger given yesterday and hit a beautiful match within the 1st 100 tries.  And got nothing else even close after 4000 tries.  I understand that the images are given randomly.  But it would be nice if the computer could somehow adapt to the close choices we make and focus in on the parameters that generated those close matches.  Maybe this isn't possible the way the applet works (remember the lack of programming knowledge
I know there are such things as computer adaptive testing that take the given answers into account when generating the next question. I just wondered if it could be applied here?

[Zeus2007]

Huh I wasn't blaming the tools.  I know how important the save option is, I use it.  It's just that sometimes I'm too taken by excitement that I even forget who I am.  Space tend to have that effect on you from time to time.

[aholinch]

... But it would be nice if the computer could somehow adapt to the close choices we make and focus in on the parameters that generated those close matches.  Maybe this isn't possible the way the applet works (remember the lack of programming knowledge
I know there are such things as computer adaptive testing that take the given answers into account when generating the next question. I just wondered if it could be applied here?

I really like your analogy to adaptive testing.  That is a good one for me to use in understanding this myself.  I'll see if I'm able to use it successfully when attempting to explain the situation to others, or if I only sound confused.

In computer adaptive testing, your answers are compared to the unique correct answer for each question.  There is a lot of sophistication on how to present the next question based on your pattern of previous correct and (hopefully not too many) incorrect answers.  But for each question, there is a single right answer known in advance by the software.

For our situation, we clearly don't know the correct answer ahead of time.  We also don't know for sure that there is only a single correct answer.  Adaptive systems work great where you know for sure there is only one correct answer, or you only need to find a single correct answer out of many possibilities.  Consider an adaptive system that users interactive evolution - feedback from a person - to design an antenna that meets certain design goals.  Once you've gotten a single antenna design that meets all the goals, you're done.  You don't need to prove that the new design is the only design that meets the goals.

In our situation, we're trying to not only determine the "correct" simulation parameters, but also how many such sets exist. 

We do have the ability to use other adaptive/evolutionary computation techniques.  However, our experience tends to show you need at least 50 plausible candidates before you can evolve them.  For a single session, that's probably asking for a lot.  For the collective set of responses from all our volunteers, that's definitely an aspect of what we'll be doing to try to teach the computer to pick good matches.

Again, good analogy.

Anthony

[MKS]

Thank you!  A very clear explanation!  OK (sigh) - back to the sims

[graham d]

I finally made it to level 3 Jules  on the current merger

[jules]

I finally made it to level 3 Jules  on the current merger

Was hard pressed to get to level 1 yesterday - so few candidate mergers. That was a tough one. Have yet to have a go at today's.

[ElisabethB]

today's one look's easy but oh boy !!!!

[graham d]

Eat your hearts out ladies . ........got to level 2 on it; half an orbit. What intrigues me is that the start point has no distortion in either galaxy- yet they must have have approached the start point from far away. I have my doubts that its a viable simulation.

[graham d]

Image inversion of Arp 72 does reveal there is a tail and bridge of star tails to the small companion.
http://cas.sdss.org/dr7/en/tools/explore/obj.asp?id=587739810496708646

The small companion doesn't show much rotational angular momentum and it appears highly unlikely that it could have induced the major rotation or perturbed it in the large barred spiral.
I can't get anywhere near a reasonable simulation. There could be a third party that departed the vicinity. Zooming out however, doesn't reveal any other candidate. Perplexing.

[mkoskim]

Hi,

I was wondering if the applet creates canditates all the time purely randomly, or does it use any sort of genetic algorithm? If it uses, it certainly does not look like one - you get huge numbers of not-so-close matches, even if you have relatively good matches stored in your simulations. That makes the usage of the applet like not getting anywhere... If it does not use any sort of genetic algorithm, could there be one? So that it would analyze the set you have selected and trying to randomly modify and/or mate those to get new better canditates.

[ElisabethB]

hi
the latest one : SDSS 587739721376202860 looks very easy, but actually it's one of the most difficult mergers to simulate.
The fact that the black and white pictures have almost no features don't make this any easier.
Almost 50 % (or more) of the sims looks like this merger and there is no way with this level of (non existing) details to make it any better
Sorry
E.