Reading and Discussion 12: Week 12 – Dealing with Scale

Initial Posting Due: Tue, Nov 21 at (Canvas Link)

Readings

This is a big and important topic, but rather than require a lot of reading, I’ll give you less – and hope that you’ll go beyond the minimum.

These 3 things are required. The Munzner chapters are fairly short, and the TSNE web page is light reading and fun to play with.

1. Reduce Items and Dimensions (Chapter 13 from Munzner’s Visualization Analysis & Design) (Munzner-13-Reduce.pdf 440 kb)
2. Embed: Focus+Context (Chapter 14 from Munzner’s Visualization Analysis & Design) (Munzner-14-Embed.pdf 538 kb)
3. How to Use T-SNE Effectively – I wanted to give you a good foundation on dimensionality reduction. This isn’t it. But… it will make you appreciate why you need to be careful with dimensionality reduction (especially fancy kinds of it).

These were going to be required. Instead, consider them “strongly recommended”.

1. Ellis, Geoffrey, and Alan Dix. “A Taxonomy of Clutter Reduction for Information Visualisation.” IEEE Transactions on Visualization and Computer Graphics, 2007, 1216–23. (pdf) (doi)
2. Chapter 3 of Alper Sarikaya’s thesis – This is a survey of different ways of doing summarization that appear in the visualization literature. There is a lot about how the survey was conducted, but the main thing for class is to see the different categories of summarization and how they interact. This is a chapter from a thesis and might be a little harder to read out of context. (We need to write a paper version of it)

Optional

• Elmqvist, Niklas, and Jean-Daniel Fekete. “Hierarchical Aggregation for Information Visualization: Overview, Techniques, and Design Guidelines.” IEEE Transactions on Visualization and Computer Graphics 16, no. 3 (2010): 439–54. (pdf) (doi)

Online Discussion

Initial Posting Due: Tue, Nov 21 at (Canvas Link)

It’s Thanksgiving week. You have Design Challenge 2 due. But, scale is a really important topic.

Consider the “most” information you can cram into a visualization. In theory, you could assign each data point to a pixel – so each pixel represents a data point. People actually do this (there are papers about pixel-oriented displays).

For two discussion postings:

1. What are the “most dense” (but still effective) visualizations you have seen? How much information can you cram into a number of pixels? (or at least what is the most that you’ve seen) What prevents us from really getting to the “data point per pixel” level?

2. Many real cases have more data points than pixels (and certainly more data than you can show conveniently). In class / the readings we talked about a few basic strategies. Give an example of using each.