A08 | Learning and Explaining Dimensionality Reduction through Visualization

Prof. Michael Sedlmair, University of Stuttgart
Email | Website

Michael Sedlmair

Prof. Daniel A. Keim, University of Konstanz
Email | Website

Daniel A. Keim

René Cutura, University of Stuttgart – Email | Website

Dr. Quynh Quang Ngo, University of Stuttgart – Email | Website

Katrin Angerbauer, University of Stuttgart – Email | Website

In recent years, machine learning has gained much attention for its ability to model complex human tasks, such as driving cars or composing music. In visualization research, there is currently a large effort to investigate how visualization can support machine learning research and practice.

In this project, we will take the reversed perspective and investigate how machine learning can support visualization research and practice. In particular, we will leverage machine learning to build and evaluate a new generation of models for visual perception and design.

Visualizing data is a process that involves many delicate design choices: How should the data be aggregated? Which visual encoding should be used? And how should it be parametrized?

In oder to make good design choices, many alternatives to aggregate and represent the data need to be evaluated. To make the work with the data more effective and easier, the project pursues several goals.

Goals

Novel models for visual perception and design decisions.

A new user-oriented research methodology.

Evaluating and characterizing the methodology.

Fig.1: Illustration of the proposed learning-based methology using class seperation as an example. This novel user-oriented testing methodology will help us in bridging quantitative and qualitative methodes.

Fig. 2: A typical perceptual task that could be modeled using our methodology is class seperation scatterplots.

Project Group A

Models and Measures

 

Completed

 

Project Group B

Adaptive Algorithms

 

Completed

 

Project Group C

Interaction

 

Completed

 

Project Group D

Applications

 

Completed