References

Program

Tools

http://antv.alipay.com

How do people use visualization?

Provenance

The steps the user take in the process of visual exploration/analysis and the resulting visualizations and findings

We use provenance for:

Capture:

Visual Interaction Techniques

A well-designed interactive visualization interface should show the following

Visualize
use effective visual encodings
Filter
reduce visible data to relevant items
Select
Retrieve details about interesting items
Navigate
Pan, zoom, change view

https://github.com/juba/scatterD3 DeepTree http://www.visualcomplexity.com/vc/

Visual Design and Encoding - Westermann

The purpose of visualization is insight, not pictures. - Schneiderman

Why use Interactivity?

  1. Handle data complexity
  2. A single static view can show only one aspect of data

Overview first, zoom and filter, then details-on-demand

Why depend on vision?

  1. Visual system is high-bandwidth channel to brain

Pre-attentive Processing

Gestalt Principles

Representations should be correct, accurate and truthful.

To bring up a change, you must attend to it. (Change blindness)

Visual Design

A good visualization depends on:

  1. data types
  2. context of the data
  3. tasks to perform e.g. identify trends
  4. questions to answer
  5. messages to deliver

\begin{equation} \text{Lie Factor} = \frac{\text{Size of effect shown in graphic}}{\text{Size of effect in data}} \end{equation}

Bad visualizations do not allow you to recover original data from the visualization. Keep proportions and relative sizes.

maximize data-ink ratio

Steven’s Psychological Power Law

https://en.wikipedia.org/wiki/Stevens%27s%5Fpower%5Flaw

Steven’s psychophysical power law:

\begin{equation} \text{Perceived sensation} = \text{Physical Intensity}^T \end{equation}

Compensating for human’s over/underestimation:

Difficult to focus on one channel when multiple channels are presented. (Redudancy is bad!)

Visual mapping - Separable vs integral visual channels

Scientific Data Visualization - Stefan Bruckner

Types of Visualization

  1. Volume Visualization
    • Visualization of scalar fields
    • Important in medicine, biology, geoscience, engineering, …
  2. Flow Visualization
    • Visualization of Vector Fields
    • Data typically from computational fluid dynamics (CFD) simulations

Data Representation

Grids

Regular Grid

Rectilinear Grid

Curvilinear Grid

Block-structured Grid

Unstructured Grid

TODO Other Grids SUMMARY OF GRID TYPES

Scattered Data

Interesting to look at dimensionality of data space, vs dimensionality of data attributes

Data Enhancement

Data, Visualization, Interaction

Interactive Steering

Volume Visualization

Challenges

Voxels vs Cells

Linear Interpolation

Evaluating Quality of Reconstruction

Classification

Visualization Approaches

Slicing
display of 2D cross sections
Indirect Volume Rendering
Extraction of an intermediate representation
Direct Volume Rendering
Direct display of representation

TODO Isosurface Similarity

Visualization in the Spatial Domain

Indirect Volume Rendering

Marching Cubes is a standard method for the extraction of isosurfaces from volume data

Flow Visualization

Data Visualization of Text Data - Jaegul Choo

Overview

  1. Vector encoding techniques of text
    1. Bag-of-words vectors and word embedding
  2. Basic text visualization techniques
    1. Word cloud, wordle, word tree, phrase nets, ThemeRiver
  3. Topic Modeling
    1. Non-negative matrix factorization
    2. UTOPIAN and visual analytic systems
  4. Dimension reduction
    1. Multidimensional scaling and tSNE
    2. Interactive dimension reduction techniques and systems
  5. Interactive visualization of deep learning
    1. Toolkits: Tensorboard, Embedding Projector, Visdom
    2. Advanced visual analytics systems: CNNVis, LSTMVis, DeepEyes