Please disable Adblockers and enable JavaScript for domain CEWebS.cs.univie.ac.at! We have NO ADS, but they may interfere with some of our course material.

Q10 Answers

• Ich verstehe die Definition vom set membership degree nicht. Vor allem der Satz . "It corresponds to set cardinality in the dual representation of sets and elements."  

• Ich finde die Table 2 mit den Stärken und Schwächen der visual categories sehr hilfreich! Sollten nämlich mal sets visualisiert werden müssen, kann man hier sehr schnell sehen, welche Art der Visualisierung man am besten wählen sollte.  

• Auf den letzten Seiten, bei den possible opportunities wird von Interaktion gesprochen. Die habe ich auch in all den gezeigten Visualisierungsmöglichkeiten vermisst. Vermutlich ist es aber auch sehr schwierig bei so komplexen Set Daten darauf einzugehen, denn die statischen Diagramme sind teilweise schon schwer nachzuvollziehen. Wenn dann noch Interaktion des Users dazu kommt wird es wohl kind of messy…  

• is weaving used in practice? I think it's awful ( section 4.1.7 )  

• in general this stuff gets very complex/large very fast - are there proposals that regulate "modelsize" of visualizations? amount of variability that can be explained by each graphic component?  

• how much work is it to come up with these visualization concepts? how fast paced are developments in vis? really amazing!  

• Why are set memberships often abstracted as separate Boolean attributes ?  

• What are separate Boolean attributes and how are they defined ( I have read this answer "Boolean attributes representing  
  the sets that can also be used to specify which elements belong to them (So which Elements ? The data ?)) ?  

• Why can Venn diagrams be drawn for any data with only two sets and not for more sets ?  

• How would the Venn diagram be affected by more sets ?  

• For the matrix based techniques: how well does this approach scale? As the dimensions become higher so does the complexity. Is there a limit?  

• For the matrix based techniques: How does one re-order the data for clustering?  

• Is there any way to transform not well-formed diagrams into well-formed diagrams?  

• The paper discusses that a pie chart is a wrong decision to visualize the set sizes with. In which cases could be pie chart a good decision or should it be generally avoided?  

• Are there some techniques for the visualization of overlapping sets, which could be used with some modifications for the visualization of hierarchies or non-overlapping groups as well?  

• Are there some general rules one can follow to figure out which similarity measure is appropriate for a given data set?  

• Is there a difference in effectiveness between area-proportional diagrams and glyphs to show cardinality?  

• KMVQL: is this system not incredibly cluttered, even with only a few sets? How and when would one use this system?  

• In section 3, there is an overview of the most common tasks regarding sets. I really liked set A and B, but C posed some problems for me. In my understanding of a set, I assume something like: 'an element x belongs to set A if it has the following properties(or attributes as they call them)'. Given the fact that we know how the sets relate to one another my first question is, doesn't this make set C of generic tasks obsolete? For example C1 (Find out the attribute values of a certain element.) would mean 'Find out to which sets the certain element belongs to', which is the same as A1.  

• Also they are talking about different ways of visualising sets. Is the general assumption that all the minimal relations between sets are known from a mathematical perspective (e.g. if we have 10 sets, do we know already exactly which intersect and which don't and so on?)? Is the visualisation of the set meant to help the eye find whether a set A is included or not in other set B, so that the person analysing the sets doesn't have to do math? Or is it more like 'we have a bunch of sets that we have no idea about and through our visualisation we will gather some information'?  

• In section 4.1 they describe Euler diagrams and techniques to improve them for the eye of the viewer. We know from class that humans are biased towards interpreting area, especially if it's a round shape (evil pie charts), however in Table 2 they present the Euler diagram's property of being well matched from an area perspective as an advantage. If the eye can't really interpret it rightly, is it still such an advantage? And are those glyphs seen as an improvement over area proportional euler diagrams or are they competing designs?  

• If I have a dataset with multiple categorical attributes, I could interpret each of them as partitioning of the data into sets. Does that mean my data is ‘set-typed’ or does it also depend on the semantics of my attributes?  

• It seems to me that sets are unordered lists elements. In the paper there was also an example about set-visualization indicates clustering within a graph. Is there a fundamental difference between sets and clusters or are those two terms to indicate the context and possible tasks for this list of elements?  

• (The next one was pretty much answered in section 6 of the paper, so if you do not want to add anything, you can scrap it.)  
  Set visualization techniques seem to be either not vary scalable or rather exotic and hard to understand. How would one go about building an application for large datasets in such a situation? (I am aware that this is largely dependent on the exact tasks, data, and users, but are there some general approaches for such a problem?)  

• How is the scalability of techniques determined? Especially since it can vary dependent on task and data.