Figure 1.
Standard visualization pipeline.
Data feeds into a mathematical model
that relies also on parameters
, and produces a visualization
. The users
make sense of the visualization to the best other their abilities. To correct any visual inaccuracies, users must either change
,
, or
.
Figure 2.
A “galaxy view” of text data created by the IN-SPIRE suite of data visualizations.
In-SPIRE uses complex mathematical models in order to discern structure (e.g., clusters) in high-dimensional data.
Table 1.
A non-exhaustive list of parametric interactions.
Table 2.
A non-exhaustive list of V2PI.
Figure 3.
The bi-directional visualization pipeline.
Step 1) Create visualization based on a mathematical model or algorithm
that depends on data
parameters
; Step 2) display the visualization for users
to assess, Step 4) Users adjust the visualization to offer model feedback; and Step 5) Update the model
(e.g., via the parameters
).
Figure 4.
Figure A displays the simulated data in three dimensions. Observations in red, green, and blue denote groups 1, 2, and 3 respectively. Figure B displays the PCA projection of the simulated data with 20 observations (that were selected at random) highlighted. Again, red and green points represent observations in groups 1 and 2 respectively. Figures C and D show updated displays after an adjustment to Figure B. Figure C is the result of moving points marked by ‘’ in Figure B apart and Figure D is the result of moving the points marked by ‘
’ in Figure B together. Notice that both adjusted visualizations capture the clustering structure.
Figure 5.
MDS view of the the city-data and an example of cognitive feedback.
Figure A displays an Initial MDS view of the data set that describes 25 cities with 10 real variables and 20 noise variables. Figure B displays an example of cognitive feedback that arranges a set cities by relative geographic locations.
Figure 6.
A visualization of the city-data that was updated by a parametric version of the cognitive feedback plotted in Figure 5B.
The updated locations of the cities were stretched and rotated to overlay on a map of the United States. The symbols and
mark true and projected city coordinates by WMDS- V2PI. The estimated and true city coordinates are close.
Figure 7.
New cognitive feedback and updated view of city-data.
Figure A plots another example of cognitive feedback that groups college towns separately from large cities. Figure B plots an updated visualization of the data that accounted for the feedback in Figure A.