Information visualization

Information visualization

1. Information visualization and cognition

   • Amplifies cognition
     • Increased resources
       • High-bandwidth hierarchical interaction
       • Parallel perceptual processing
       • Offload work from cognitive to perceptual system
       • expanded working memory
       • expanded storage of information
     • reduced search
       • Locality of processing
       • high data density
       • spatially indexed addressing
     • enhanced recognition of patterns
       • Recognition instead of recall
       • Abstraction and aggregation
       • Visual schemata for organization
       • Value, relationship, trend
     • Perceptual inference
       • Visual representations make some problems obvious
       • Graphical computations
     • Perceptual monitoring
     • Manipulable medium

2. Data types

   • 7 basic datatypes
     • 1,2,3-dimensional
     • multi-dimensional
     • temporal
     • tree
     • network
   • variables
     • qualitative
       • categorical
       • ordinal
     • Quantitative
       • interval
       • Ratio

3. User goals and tasks

   • High level
     • Explore
     • analyze
     • browse
     • assimilate
     • Triage
     • assess
     • Understand
     • compare
   • low level
     • Retrieve value
     • Filter
     • Compute derived value
     • find extremum
     • sort
     • determine range
     • Characterize distribution
     • Find anomalies
     • Cluster
     • corrolate

4. Design

   • Use knowledge of perception and cognition to make effective visualizations
     • Gestalt laws
       • proximity
       • Similarity
       • Continuity
       • Symmetry
       • Closure
       • Relative size
       • Figure and ground
     • We sample the world through visual queries
     • Our visual working memory has very limited capacity
     • What stands out
       • Color
         • Surrounded colors do not pop up
       • Orientation
       • Motion
       • Size
       • shape
       • Number
       • Enclosure
       • Addition
       • Parallelism
       • Shading
       • Depth of field
       • Compound features
       • Laws of pre attentive display
         • must stand out on some simple dimension
           • color
           • Simple shape
             • orientation
             • size
           • motion
           • depth
         • Pre-attentive channels
           • pop outs can be created by
             • form
             • color
             • simple motion/blinking
             • Addition
             • Number (up to 3)
             • Spatial, stereo depth, shading, position
         • pre-attentive conjuctions
           • color and motion
           • color and position
           • shape and position
           • Spatial location and some aspect of form
   • Use design guidelines for developing effective visualizations
     • Multiview design
       • Cost-benefit tradeoffs
         • cognitive aspects
           • the effort required for context switching
           • the effort required for comparison
           • the load on the user's working memory
           • The time and effort to learn the system
       • The Data
         • entirely different information
         • Aggregates of individual values
         • Dataset as subset of another
       • Rules
         • Diversity
           • ◦ Use multiple views when there is a diversity of attributes, models, user profiles, levels of abstraction, or genres.
         • Complementarity
           • Use multiple views when different views bring out correlations and/or disparities
         • Decomposition
           • Partition complex data into multiple views to create manageable chunks and to provide insight into the interaction among different dimensions.
         • Parsimony
           • Use multiple views minimally
         • Space-time resources
           • Balance the spatial and temporal costs of presenting multiple views with the spatial and temporal benefits of using the views
         • Self-evidence
           • Use perceptual cues to make relationships among multiple views more apparent to the user
         • Consistency
           • Make the interfaces for multiple views consistent, and make the states of multiple views consistent
         • Attention management
           • Use perceptual techniques to focus the user's attention on the right view at the right time.
     • 8 Visual variables
       • Motion
       • Texture
       • Orientation
       • Color
       • brightness
       • Size
       • mark/shape/glyph
       • position
     • Clutter reduction
       • Technique catagories
         • Appearance
           • sampling
           • Filtering
           • Change point size
           • Change opacity
           • Clustering
         • Spatial distortion
           • point-line displacement
           • topological distortion
           • Space filling
           • Pixel plotting
           • Dimensional reordering
         • temporal
           • Animation
       • Criteria to characterize methods
         • Avoids overlap
         • Keeps spatial information
         • Can be localized
         • Is scalable
         • Is adjusted
         • Can discriminate points/lines
         • Can see overlap density
     • Aggregation
       • types
         • Hierarchical
       • statistical aggregations
         • count/sum
         • average
         • Mode
         • extents
         • median
         • percentiles
         • distribution
       • Design guidelines
         • Entity budget
         • visual summary
         • Visual simplicity
         • Discriminability
         • fidelity
         • Interpretability
     • Scented widgets
       • Defenition
         • enhanced user interface widgets with embedded visualizations that provide information scent cues for navigating the information space
       • Scent encoding guidelines
         • All widgets visualizing the same scent data should use matching visual encodings
         • Modes of encoding should reinforce semantic relationships between the widget scent and encodings in the application
         • Visualizations show the same scent data should be scaled indentically across all widgets
         • Modes of encoding should respect existing interface conventions
         • Encodings which make some elements markedly more salient than others, such as opacity should be used with discretion
       • Layout guidelines
         • Interfaces should be laid out so that scented widgets are sufficiently proximal to allow comparisions between them
         • Scented widgets should be grouped, sized, aligned, and oriented similarly in order to provide common axes on which to compare scent
       • Composition guidelines
         • The overall number of type of scented widgets in a given interface should be small enough to allow easy comparison and visual tracking of changes
         • Widgets should include identifiers (icons, tooltips, text etc) that indicate what the scent cues correspond to.
   • Interaction models
     • defines how interactions are related to one another
     • Interactive visualization
       • High level
         • Interaction catagories Yi et
           • Select
             • provide ability to mark a data items of interest to keep track of
           • Explore
             • Enable users to examine a different subset of data cases
           • Reconfigure
             • Provide users with different perspectives on the data by changing the spatial arrangements of representations
           • Encode
             • Changing the type of representation to reveal new insights
           • Abstract/elaborate
             • Provide users with the ability to adjust the level of abstraction of a data representation
           • Filter
             • Enable the user to change the set of items being represented based on some specific conditions
           • Connect
             • Highlighting of associations or relationships between data items that are already represented or show hidden items that are relevant to a specified element

5. Process

   • Set up information visualization evaluation schemes
     • Evaluation
       • Many different views
         • Usability and Aesthetics
           • Qualitative experiments
         • The quality of the result
           • benchmark results
         • User effort
           • number of interactions
           • amount of eye movements needed
         • Value of visualization
         • insight gained

6. Algorithms

   • Data transformation
     • Dimensionality reduction
       • linear
         • PCA
       • non-linear
         • Probabilistic
           • SNE
         • Distance preservation
           • MDS
         • Kernel based
           • Kernel PCA
         • Preserving local properties
           • LLE

7. Visual analytics

   • Models and frameworks

8. Text analytics