Visualization M. Behrisch , M. Blumenschein, N. W. Kim, L. Shao, M. - PowerPoint PPT Presentation

Visual Computing Group, Harvard for Information Visualization M. Behrisch , M. Blumenschein, N. W. Kim, L. Shao, M. El-Assady, J. Fuchs, D. Seebacher, A. Diehl, U. Brandes, H. Pfister, T. Schreck, and D. Weiskopf, D. A. Keim

Visual Computing Group, Harvard Effectiveness Usefulness Appropriateness Efficiency Usability Expressiveness Expressiveness Interpretability

Visual Computing Group, Harvard Patterns in Information Visualizations NO NO (NO) (NO) OK OK Conceptual/ Task Task Pattern Space 3

Visual Computing Group, Harvard Patterns in Information Visualizations NO NO (NO) (NO) OK OK Conceptual/ Task Task Pattern Space 4

Visual Computing Group, Harvard Patterns in Information Visualizations Log Lin Log Lin User Visualization Visualization Dataset Understandability Parameter Parameter Characteristics 5

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) D U T 𝜚 min arg max 𝜚 ∈ Φ Quality Optimization Criterion Algorithm

Visual Computing Group, Harvard Structure and Goals of the Survey Research Goals Multi- and High-Dimensional D 1 D 1 Dim1 D 6 D 2 D 6 D 2 Reference Manual for QM Dim2 D 5 D 3 D 5 D 3 Dim3 Establish Common Vocabulary D 4 D 4 Relational Data Geo-Spatial Data Open Challenges and Future Research Directions Sequential/Temporal Text Data 7

Visual Computing Group, Harvard Structure and Goals of the Survey Research Goals Multi- and High-Dimensional D 1 D 1 For each Vis Type 134 Paper Dim1 D 6 D 2 D 6 D 2 Reference Manual for QM Selection Dim2 … D 5 D 3 D 5 D 3 per Vis. Type Dim3 1. Visualization Description In total ~300 Paper Establish Common Vocabulary D 4 D 4 Relational Data Geo-Spatial Data 2. Why do we need QMs? Open Challenges and Future A B C 14 Vis Types 3. Typical Analysis Tasks A B C A B C Research Directions Categorization A B C 4. Summary of Approaches per Vis. Type 5. Evaluations Methods Sequential/Temporal Text Data 6. Open Research Questions A B C Insight Generalization Independent of Vis. Types 8

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) min arg max 𝜚 ∈ Φ Clutter Removal vs Pattern Retrieval

Visual Computing Group, Harvard Auto-Sampling – Clutter Removal Parallel Coordinates Optimization Overplotted% Percentage of pixels containing Patterns and Tasks more than one plotted point G rouping C orrelation O utlier T rend Overcrowded% Percentage of plotted points hidden behind a pixel Hidden% Percentage of plotted points that are hidden due to being overplotted [Ellis2006]

Visual Computing Group, Harvard Scagnostics – Pattern Retrieval Convex Alpha Minimum Scatter Plots Hull Shape Spanning Tree Optimization a Convex: Area of Alpha Shape v b divided by area of Convex Hull v Skinny: Patterns and Tasks v Ratio of perimeter to area of the G rouping C orrelation v O utlier T rend Alpha Shape v [Dang2014] Stringy: v Ratio of 2-degree V in MST to # of V > 1-degree [Wilkinson2006] 11

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) min arg max 𝜚 ∈ Φ Analysis Scenarios/Tasks for QM Clutter Pattern-Driven Search for Data Groups Removal Analysis Search for Outliers Search for Dimension Relations Reduces Focuses on Cognitive Overload Analysis Task Preservation Task Data- and Visualization Specific Tasks

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) min arg max 𝜚 ∈ Φ Explicit and Implicit QM

Visual Computing Group, Harvard Noise Dissimilarity – Explicit QM Pixel-based Techniques Optimization Line Hilbert Spiral Patterns and Tasks G rouping C orrelation O utlier T rend [Albuquerque2011] 14

Visual Computing Group, Harvard Force Directed Layout – Implicit QM Node-Link Diagrams Bends Optimization Edge crossings M. Kaufmann und D. Wagner, editors. Drawing Graphs — Methods and Models . Springer, 2001 Minimum Angles Input: Graph G = (V,E) Start with random placement of vertices Orthogonality Repeat k times (k constant){ Patterns and Tasks 1. for all v in V do Calculate the repelling forces on v G rouping I tem G roup P ath Symmetry that are excerted by V \ v 2. for all e = (u, v) in E do Calculate the attracting forces between u and v [Purchase2002] Connectivity Connectivity Connectivity Cluster Overlap + Cluster Overlap + 3. for all v in V do Add repelling and attracting forces Cluster Overlap Cluster Overlap + Circle + Star Constraint Star Constraint [Wang2018] Move v in direction F(v) Circle Constraint } 15

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) min arg max 𝜚 ∈ Φ Data Space vs Image Space

Visual Computing Group, Harvard Quality Metric q( 𝜚 | D, U, T) min arg max 𝜚 ∈ Φ Quality-Metrics -Driven Automation QM Influence Data Specification Computation Aspects User / Task Concept Intuition Biases Adaption Evaluation View Source Data Transformed Visual Visual Transformation Views Data Data Structures Transformation Mapping User Rendering Data Space Image Space Quality Metrics Quality Metrics [adapted from Bertini2011]

Visual Computing Group, Harvard TextFlow – Data Space QM Stacked Charts Optimization [Cui2011] Patterns and Tasks P arallel T rend S plit / M erge Topic Flow Topic Bundles Thread 18

Visual Computing Group, Harvard Magnostics – Image Space QM Matrix Optimization [Behrisch2016] C1 C1 Pattern Response Patterns and Tasks C2 C2 Pattern Variability G rouping I tem G roup P ath C3 C3 Pattern Sensitivity C4 C4 Pattern Discrimination Negative Connectivity Connectivity Connectivity Example Low High FUZZY_ OPPOaNENT_ Distance Distance HISTOGRAM 19

Visual Computing Group, Harvard Quality Metrics Landscape High-Level Meta-Perception / User Cognitive Process / Complexity Clutter Removal vs Pattern Retrieval Memorability, Understanding, Confidence, Faithfulness, Trustworthiness, Cognitive Biases, Engagement, User Level, Conventions, Aesthetics, Joyfulness Mid-Level Perception / Task Explicit vs Implicit QM Patterns versus Anti-Patterns , Clutter-reduction, Task-effectiveness “Overview -First & Details-on- Demand”, “Search, show context, expand on demand” Data Space vs Image Space Low-Level Perception Preattentive Processing, Gestalt Laws, Visual Variables, Change Blindness, Just-Noticeable-Differences 20

Visual Computing Group, Harvard Quality Metrics Landscape High-Level Meta-Perception / User Color Research Cognitive Process / Complexity Memorability, Understanding, Confidence, Faithfulness, Trustworthiness, Cognitive Biases, Engagement, User Level, Conventions, Aesthetics, Joyfulness Mid-Level Perception / Task Patterns versus Anti-Patterns , [Gramazio2016] [Mittelstädt2015] Clutter-reduction, Task-effectiveness “Overview -First & Details-on- Demand”, 39 studies about human perception in 30 minutes “Search, show context, expand on demand” Low-Level Perception Preattentive Processing, Gestalt Laws, Visual Variables , Change Blindness, Just-Noticeable-Differences [Elliott2016] 21

Visual Computing Group, Harvard Quality Metrics Landscape High-Level Meta-Perception / User Cognitive Process / Complexity Memorability , Understanding, Confidence, Faithfulness, Memorability Trustworthiness, Cognitive Biases , Engagement, User Level, Conventions, Aesthetics, Joyfulness [Borkin2015] Mid-Level Perception / Task Patterns versus Anti-Patterns , Clutter-reduction, Task-effectiveness “Overview -First & Details-on- Demand”, “Search, show context, expand on demand” Aesthetics Low-Level Perception Joyfulness Preattentive Processing, Gestalt Laws, Visual Variables, Change Blindness, Just-Noticeable-Differences [Skau2017] 22

Visual Computing Group, Harvard Quality Metrics Landscape High-Level Meta-Perception / User Mid-Level Perceptual Quality Metrics Cognitive Process / Complexity Memorability, Understanding, Confidence, Faithfulness, Trustworthiness, Cognitive Biases, Engagement, User Level, Conventions, Aesthetics, Joyfulness Clutter Pattern-Driven Removal Analysis Mid-Level Perception / Task Reduces Focuses on Patterns versus Anti-Patterns , Cognitive Overload Analysis Task Clutter-reduction, Task-effectiveness “Overview -First & Details-on- Demand”, “Search, show context, expand on demand” Computation User / Task Data Spec. Concept Aspects Intuition Low-Level Perception Quality Metric Preattentive Processing, Gestalt Laws, Visual Variables, Influence Change Blindness, Just-Noticeable-Differences 23

Visual Computing Group, Harvard Discussion and Findings T Which QM favors which visual pattern? Quality Metric > Implicit, domain-inspired, pot. subjective expectation q( 𝜚 | D, U, T) min arg max > What if pattern is not known apriori? Which QM? 𝜚 ∈ Φ > Majority of QMs do not describe visual pattern 24

Visual Computing Group, Harvard Discussion and Findings OLO What are extreme cases that a QM can deal with? PCA Sloan Quality Metric q( 𝜚 | D, U, T) min arg max > Noise (in-)variances and robustness toward 𝜚 ∈ Φ skewed distributions > Bad QM must mean no pattern 25

Visual Computing Group, Harvard Discussion and Findings Is QM research transferable among visualization types? [Fink2013] [Heer2006] Quality Metric > Some vis subdomains share similar concepts q( 𝜚 | D, U, T) min arg max > Set of base patterns in both visualizations 𝜚 ∈ Φ 26