Loizides, Andreas;
(2003)
Intuitive visualisation of multi-variate data sets using the empathic visualisation algorithm (EVA).
Doctoral thesis (Ph.D), UCL (University College London).
![[thumbnail of Intuitive visualisation of multi-variate data sets using the empathic visualisation algorithm (EVA).pdf]](https://discovery.ucl.ac.uk/style/images/fileicons/text.png) |
Text
Intuitive visualisation of multi-variate data sets using the empathic visualisation algorithm (EVA).pdf Download (10MB) |
Abstract
The central thesis of this research is that there exists an algorithm that can produce a naturalistic visual structure (such as a human face) that represents a multivariate data set that can be utilised to discover hidden features in the data. Research in this thesis lies in the area of Information Visualisation and is concerned with techniques for visualising large scale multivariate data sets in order to help in understanding and exploration. Examples include financial data, business information and results from experiments. A visualisation method maps such data ideally into intuitive visual structures, however, in most cases there is no obvious mapping from the data to the visual structure. This thesis explores the use of naturalistic visual structures (for example, human faces) as representations of such multivariate data sets. An automatic mapping from the data to the visual structure is constructed through the use of the empathic visualisation algorithm (EVA) implemented for this purpose. EVA, is a fundamental extension of the type of data visualisation first introduced by Chernoff, who exploited the idea that people are hardwired to understand faces and therefore can quickly interpret information encoded into facial features. We use faces as our paradigmatic example, but the method is not limited to this case only. Given an n x k data matrix of n observations on k variables, the original Chernoff method assigns each variable to correspond to a particular facial feature like shape of the nose, or shape of the eyes. The mapping from data to visual structure is arbitrary, and the resulting faces have no correspondence to the underlying semantics of the data. Such faces are good for understanding pattern, but any individual face seen in isolation does not readily convey anything about the data without knowledge of the specific mapping used. EVA provides an automatic mapping from semantically important features of the data to emotionally or perceptually significant features of the corresponding visual structure, such as a human face. In other words a single glance at the visual structure informs the observer on the global state of the data, since the visual structure has an emotional impact on the observer that is designed to correspond to the impact that would have been generated had the observer been able to analyse the underlying data itself. Finer details concerning interpretation of the visual structure are then available through knowledge of the relationships between semantically important features of the data and emotionally significant aspects of the visual structure. EVA uses a Genetic Program (GP) to map the quantitative measurements from the multidimensional data set to the qualitative measurements of the visual structure. The genetic program typically converges after about 75 generations. The experimental data supports the main thesis of this research.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Intuitive visualisation of multi-variate data sets using the empathic visualisation algorithm (EVA) |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Thesis digitised by ProQuest |
| Keywords: | Applied sciences |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10100739 |
Loading...
Loading...Loading...Loading...Archive Staff Only
 |
View Item |
