Help with merger enhancement tool

[skylark42]

With the new merger enhancement tool - which I tried on Tuesday - the appearance of the galaxy pair is extremely sensitive to the exact position of the cursor as you move it about in "speed", "depth", "mass". So when you're close to such an explosive "node", things change hugely for a very tiny movement of the cursor. To get the best match, you sometimes need to be able to make much smaller movements than you can with the mouse. But my arrow keys don't seem to work. Is there a way around this?

[Hanny]

(Just to let you know I've deleted the double topic, which was posted by accident, I'm afraid I don't know the answer to your question). Welcome to the Zoo!

[jwallin]

Hi all,

The merger enhancement tool IS tricky to use.   The systems we are modeling are highly non-linear, so tiny movements can cause tiny tiny changes or very big changes, depending on the part of phase space you are exploring.  We looked at a lot of different options to make this work better, but so far, the best method has eluded us.    We did play around with a lot of options, and didn't come up with an optimal system.   It might be a small comfort, but I believe that some of the extreme sensitivity is unavoidable because we are dealing with orbits and galaxies.

We are going to run with the current system for a while, and then look at improvements in a month or so.  Frankly, we are afraid to mess with things too much since things are currently running.  I am sorry we can't do better on the update cycles, but we will work on it and see what we can come up with for improvements based on your comments. 

Thanks so much for your patience and help!

Happy Thanksgiving!

-John

[Hanny]

Good to have you on the forum John!

[Zeus2007]

I think this refers to the little yellow cross and I don't think my mouse likes it.  Sometime I move it slowly towards the right and the cross just disappears only to reappear at the top.  I'm still on it, the project is just great.

[flashingsteel]

Cant you just make the box bigger (more resolution so each increment has a smaller effect)?  This wouldn't require many changes, and could make the controls less touchy. 

[Rgoldsb]

I just checked out galaxy zoo for the first time today and this issue immediately became apparent.  The easiest solution to me would seem to be including some sort of manual input.  Obviously units are not relevant, but you can simply define the axes to range from 0 to 100, and allow users to enter two values to specify a point in 2d space for each parameter, mass, size, etc (the graphical interface could be used as well for rough tuning).  The point could then be defined as accurately as you want, and fine tuning would simply involve tweaking smaller and smaller values.  i.e. once you have it to within 75-76, then play around with 75.5, or 75.4 etc.  If you settle on 75.4 then try 75.42,75.43, and so on and so on.  Of course this also depends on what sort of inputs the simulation is taking.  I suspect the cursor seems to jump around because the parameter space is actually divided into a discrete grid of acceptable inputs, in which case this solution could not be implemented.  If this is the case, then perhaps some left, right, up and down clickable buttons that allow you to move the cursor in the smallest possible increments would help.

[dcortesi]

I'll second the call for key input. There are fascinatin' places where the tiniest change of e.g. velocity switches the sim from one-orbit boring, to 3- or 4-turn do-si-do and an explosion of stars - and the next tiny change is back to boring. Obvious nonlinearity, to the verge of chaos!

To explore these "hot spots" with the mouse is very difficult and frustrating. It would be much nicer if the arrow keys were an alternative to the mouse. Exactly as in a drafting program, where you can drag graphic objects around, but to get them juuuuust where you want them, you go to the one-pixel "nudge" with the arrow keys. So: give us "nudge" keys!

[aholinch]

Great suggestions in here.     We definitely would like to provide some type of fine-tuning capability.  Not all of the 12 adjustable parameters are quite as sensitive as velocity.

The current cursor is meant to be smooth in that there is no discretization.  It allows you to explore the full range of values for the current parameters.  This full range is limited to certain values set based on observational data.  So you don't get plus or minus infinity.  Part of the jumpiness is due to the actual non-linearity of the response to changes in parameters.  Another part of it is that there is a finite time to calculate the simulation as you drag the cursor.

We had tested previous versions that allowed typed input for all simulation parameters.  I can tell you that as someone who has been analyzing orbits of various types for over a decade, it was a little intimidating.  There still might be a way to accommodate this though.

At the very least some type of "nudge" sounds like it would be used.  Thanks for the great suggestions everyone.

[Thomas J]

I agree, a nudge feature would be beneficial. That way, one could move the cursor in very small increments using the arrow keys. I think this would feel comfortable to use.

Oh, and as it's almost xmas, can we have full screen with gas and dust, please?  (kidding)

[graham d]

We see separate spiral or disk galaxies interact, come together and merge to form a bewildering variety of mergers. This could be a lame analogy but if you could imagine a phase space of these 12 parameters there would be many paths along which the separate galaxies evolve and change to ptoduce the merger we see and not just a single final end point, say an elliptical galaxy with random star orbits. Those paths might following an undulating topography. Some parameter value changes might be so sensitive that they cause a rapid change in undulations. It would be like walking through a mountainous region with sharp peaks and chasms.
 
If the fine tuning is so great this would lead to an intermediate solution that takes you to the very top of the peak where chaos occurs or the final depth of the cavern or chasm where chaos resides again. For a sharp peak there would have to be a rope ladder that connects to either an adjacent peak or more undulating terrain. You can't retreat back down the peak, the time function brings both galaxies to closer orbit.

Whenever I reach a simulation point where one a parameter change or aslight nudge of the cursor leads to a chaotic result I tend to think the actual merger didn't take this route. Hence I retreat and search for a different path along more undulating regions. A supersensitive one pixel nudge takes you higher to the sharp peak or deeper into the chasm where only a thin roap bridge could save you.

More often than not the velocity function becomes supersensitive. For myself I tend to think that an incremental slight change shouldn't cause such an immense change towards chaotic behaviour, otherwise we would never see common intermediate and diverse merger patterns. More sensitivity will give more detailed topographic maps that become more mathematical than physical. I would think we actually want a picture as simple as possible, with as few parameters as possible, that doesn't need so sensitive a fine tuning. Whenever I reach a point in the enhancing where one parameter appears to trap the simulation- any change leads to a chaotic pattern, I stop. Had the merger taken this route it would need the benefit of lots of rope ladders to save it. Hence, I see the appearance of sharp peaks and chasms as no go areas to avoid.

So I seek routes that change more smoothly. Some routes end up with the cursor for one parameter jammed up against the edge of the screen, at the limit of its range in two ordinates. I retreat again.

[jwallin]

Hi all,

Just a couple of updates -

1) We are looking at possible enhancements on the merger tools to make the fine tuning easier.   It is near the top of our list for enhancements right now.   We hope to have it "soon".   We are also looking at some other enhancements that will be fun.   (Yes... this is vague... but stay tuned!)

2) We have some pretty cool galaxies in the staging area right now.   We should be putting them over the next few weeks.   

3) So far, things are going very well.   I hope to do a longer blog post with the stats  in a few days.   

Thank you for all your help.  Your clicks, your support, and your time are greatly appreciated.

Zooites rock!

-John