Visual Analytics - Non-Linear Dimension Reduction (2)
In the second part of this lecture, we discuss how the resulting projections can be assessed. Moreover, we discuss how the user can affect the projections. We also discuss a whole workflow of assisted dimension reduction. This relates back to ideas discussed already in the first lecture: to provide guidance and assistance to support analysts in using visual analytics techniques. It turns out that there is not just one best way to perform dimension reduction - there are valid alternatives. Some of them better preserve outliers, others better preserve clusters or indicate correlations. Thus, using a sequence of possible dimension reduction is more likely to fully understand the data. Chapters: 00:00 - Assessment of Projection Quality 18:11 - Assisted Dimension Reduction 44:26 - Guidance 53:00 - Summary, Outlook and References
In the second part of this lecture, we discuss how the resulting projections can be assessed. Moreover, we discuss how the user can affect the projections. We also discuss a whole workflow of assisted dimension reduction. This relates back to ideas discussed already in the first lecture: to provide guidance and assistance to support analysts in using visual analytics techniques. It turns out that there is not just one best way to perform dimension reduction - there are valid alternatives. Some of them better preserve outliers, others better preserve clusters or indicate correlations. Thus, using a sequence of possible dimension reduction is more likely to fully understand the data. Chapters: 00:00 - Assessment of Projection Quality 18:11 - Assisted Dimension Reduction 44:26 - Guidance 53:00 - Summary, Outlook and References