Explore Galaxy Zoo Classifications
This post (and visualization) is by Coleman Krawczyk, a Zooniverse Data Scientist at the ICG at the University of Portsmouth.
Today we’ve added another new tool for visualizing Galaxy Zoo, this time showing the full vote path of all users for each galaxy from GZ2 onward. The first node of the visualization shows an image of the galaxy and each of the other nodes represents the answer to a question from the Galaxy Zoo decision tree, and the size of the node is proportional to the number of votes for that answer. The maximal vote path is highlighted and also shown in words across to top of the tree, and the results of the “Is there anything odd?” question are shown across the bottom.
The full Galaxy Zoo catalog can be searched via Zooniverse ID (the same one used for Talk), RA and Dec, or randomly. After picking a galaxy the nodes can be moved around by clicking and dragging, and the links can be collapsed/expanded by clicking the attached nodes, both of these functions are useful for untangling complex trees. Various properties of the visualization can also be controlled with the sliders below the tree. For a guided tour of this tool click the “Take a tour” button, and for a full list of features click the “Help” button.
Screenshot of the Visualisation Tool
Galaxy Zoo 
Karen,
A VERY interesting tool. A couple of points that I don’t understand though and think that one or two others might have an issue with:
(i) If it is GZ2 onward – presumably to the current iteration – isn’t there an issue that people will look at this and will vote along the “maximal” line, making a mess of the statistics?
(ii) Could you please set out more clearly what terms like “friction” etc mean in this context. I have taken the tour and find myself more confused than I was before.
Thanks
..to be explicit could you clarify “charge”, “friction” and “link strength” in this context and explain their significance.
Thanks
The three sliders at the bottom control the various parameters of the d3js force layout used for the visualization (see https://github.com/mbostock/d3/wiki/Force-Layout for full descriptions). The basic idea behind them are:
Charge: How much the nodes repel each other. The larger the charge the more the nodes are pushed apart. The charge on each node is also controlled by the size of the node, the slider just changes a global multiplication factor.
Link Strength: How stretchy the links connecting the nodes are. 0 mean “act like the links are not there at all,” and 1 means “act like a strong rope is connecting the nodes together.”
Friction: How strongly the node movement is damped each time step. 0 means there is no damping, 1 means 100% damping. Note, this is the opposite of how it is in the d3js documentation. I flipped it around so 0=”frictionless” instead of “full friction.”
Coleman,
I think I might have expressed it better perhaps.
I have read the github pages and I don’t see why users might want to access to these parameters. Put another way why would I want to see other than a default view that shows me the links? I can see the point of showing me what the relative scores are and showing a maximal set of links, but I really don’t get the point of the sliders.
Can you explain?
Peter
The three sliders at the bottom control the various parameters of the d3js force layout used for the visualization (see https://github.com/mbostock/d3/wiki/Force-Layout for full descriptions). The basic idea behind them are:
Charge: How much the nodes repel each other. The larger the charge the more the nodes are pushed apart. The charge on each node is also controlled by the size of the node, the slider just changes a global multiplication factor.
Link Strength: How stretchy the links connecting the nodes are. 0 mean “act like the links are not there at all,” and 1 means “act like a strong rope is connecting the nodes together.”
Friction: How strongly the node movement is damped each time step. 0 means there is no damping, 1 means 100% damping. Note, this is the opposite of how it is in the d3js documentation. I flipped it around so 0=”frictionless” instead of “full friction.”
Coleman,
See my reply of 1st May. You don’t explain why it is in the UI. Why might I want to vary any of these? Do they have any scientific meaning or is it just a question of changing what it looks like?
Sorry for the late reply, only just saw your post (the one on May 1st). The reason to have the sliders is the app will adjust its size depending on the size of the web browser window, and when the screen is small all the values need to be lowered to fit the visualization on the screen. Try viewing the site on a smart phone and you will see what I mean.
Thanks, that clarifies it.
Hi Peter,
I don’t think we’re worried about it for ongoing classifications, for two reasons:
– this isn’t on live data (doesn’t contain the live GOODS/CANDELS sets in as of April 2015, for example)
– volunteers don’t control which subjects they classify; everyone is still randomly shown an image on GZ.
– Even if we connected it to live subjects, users would need to exit the task, open the classification viewer separately, and look at what the results were. I think the chances of that are very small, but we’d simply ask the volunteers explicitly not to do it in that case.
Thanks Kyle.
Peter
I think this subject that Peter is trying to clarify totally makes sense. I, as well as Peter, needed some clarification when I was reading this article…. but thanks Kyle. Your explanation helps my understanding.