As the amount and complexity of available information grows, it becomes clear that traditional data analysis methods are insufficient. In particular, the data analysis process becomes inherently iterative and interactive: i) users start analysis with a vague modeling assumption (expressed often as a form of domain knowledge) about the data; ii) data are analyzed and the intermediate results are visually presented to the users; iii) users revise modeling assumptions and the process iterates. This process has emerged as a prominent framework in many data analysis application areas including business, homeland security, and health care. This framework, known succinctly as visual analytics, combines visualization, human computer interaction, and statistical data analysis in order to derive insight from massive high dimensional data.

Many statistical learning techniques, for instance, dimensionality reduction for information visualization and navigation, are fundamental tools in visual analytics. Addressing new challenges -- being iterative and interactive -- has potential to go beyond the limits of traditional techniques. However, to realize its potential, there is a need to develop new theory and methodology that bridges visualization, interaction, and statistical learning.

The purpose of this workshop is to expose the NIPS audience to this new and exciting interdisciplinary area and to foster the creation of a new specialization within the machine learning community: machine learning for visual analytics.