-- KimberlyDextrasRomagnino - 2016-05-03

Summaries of Research Papers that I have read

Design Studies

Relex: Visualization for Actively Changing Overlay Network Specifications
-Michael Sedlmair, Annika Frank, Tamara Munzner, Andreas Butz

May 4, 2016

Design Study

Focus on: Visual network analysis/Network visualization: building engineered networks rather than discovering evolved networks

Contributions:

• characterization and abstraction of tasks and data in this problem domain and their relationship to abstractions used in other network visualization application areas
• carefully justified and validated design of RelEx, created with a full user-centered design process

Target Group/Domain Expert: automative engineers at BMW
Tasks: specify and optimize the communicaiton wtihin in-car networks which connect the vehicle's electronic control units (ECUs) and enable the exchangeof information betwen them via several communication bus subsystems

Design Process:

• Problem Characterizaton Phase
  • Steps:
    • current practices, needs, challenges
    • analysis of success and failure factors of previously proposed tools
    • Task and Data Abstraction
      • Data: multiple related networks that are explicitly changed by the user, with a logical specification of communication paths layered on top of a physically isntantiatedbearer network
        • not node scalability but path scalability (too many edges)
        • actively changing data
        • overlay communication network whtere a lofical layer expressing communication paths is mapped onto a physical later with physcial edges
      • Tasks: specifying and optimizing traffic flow.
        • traffic optimization on the signal path network
        • queries about relations between the elements adn networks
        • focus on finding high-traffic edges instead of highly central node
        • Subtasks:
          • lcoation hotspots
          • finding number of communication partners for a specific ECU
          • characterizing the communcatio structures between entire bus sytems on the signal path network
          • determining whether a signal is well-directed between just two ECUS or widespread
    • Design Requirements
      • Integrate with existing tools
      • Extend rather than displace current practices
      • Support some aspects of the task through visual encoding, others through interaction
      • Rich exploratino of complex relations
      • Signal Paths need an explicit visual represenation
      • logical Overview is needed
  • Methods:
    • contextual inquiry
    • semi-structured interviews
    • focus group
• Desgn and Formative Evaluation
  • Four Steps ( design principles of parallel prototyping):
    • diverging phase: broad set of alternative data abstraction and visual design ideas
    • paper mockups and discussion with users
    • Refined paper prototype of one final design concept
    • implemented refiened concept as software system using agile development
• Summative Evaluation
  • Study #1:
    • field study
    • think-aloud study
      • qualitative comparison of Relex with four previous tools

Tool: Relex (Relation Explorer)

• multiple coordinated views with linked selection and highlighting
  • Matrix View: custom adjacency matrix representation for overview
  • SignalPath View: filtered detail view. node-link
  • Topology view: node link, automatic layout
  • Two List Views
• Brushing and Linking
  • domain conventions for colors
  • highlighting edges = increasing width
  • hover adn selection highlights element and close elements in all views

Notes

• Most network visualization design studies focus ont he abstract tasks of cluster analysis and the identificaitn of central nodes and outliers. Framed as the identificaiotn adn escription of communities, key actors, and atypical actor roles
• Matrix vs node-link encoding
  • Node link: poor choice for large dense graphs because they become extremely cluttered
  • Matrix: scales well even for dense graphs but is sbad at path tracing
• focus on using encodings used in previous tools to aid learnability and usability