While Scientific Visualization techniques are used for the clariﬁcation of well-known phenomena, Information Visualization techniques are used for searching for interesting phenomena. There are therefore important differences between the two fields of research. The two ﬁelds of research are compared, but since this paper is meant for Scientific Visualization researchers focus will be on explaining Information Visualization techniques. This paper list the main characteristics of the two fields of research, describes common Information Visualization techniques and discusses differences and similarities in the software that commonly is used in the two fields of research.
1. Introduction （引言）
Some researchers mainly use visualization for ﬁnding interesting phenomenon in completely unknown data, whiles others use visualization for the conﬁrmation or rejection of hypotheses. These two scientiﬁc communities live separate lives, with very little sharing of knowledge between them. While there naturally are important differences between the two kinds of visualization, it is the hypothesis of this paper that an understanding of both of these two ways of using visualization potentially could lead to new ways of visualizing data for the beneﬁt of both communities. This article therefore contains a summarization of the most important and irreconcilably differences between the ways of thinking. However, since this article is presented at a Scientiﬁc Visualization symposium, the focus will here be on Information Visualization ideas and techniques.
2. Principal Differences （主要差异）
Card et al.  deﬁnes the two forms of visualization as:
- Scientiﬁc Visualization: the use of interactive visual representations of scientiﬁc data, typically physically based, to amplify cognition.
- Information Visualization: the use of interactive visual representations of abstract, non-physically based data to amplify cognition.
While Scientiﬁc Visualization covers accurate visualizations of the real world, Information Visualization covers visualization of concepts that often are abstract in nature. The purpose of Information Visualization is to make it possible for analysts of data to obtain internal mental models of the information content in datasets;models which subsequently can be used for characterization, prediction, and/or decision making.
Keim et al.  classiﬁes visualizations into the categories:
- Exploratory Analysis: the search for hypotheses
- Conﬁrmatory Analysis: the conﬁrmation or rejection of hypotheses
- Presentation: the presentation of facts that are ﬁxed a priori
2.1 Information Visualization and Exploratory Analysis （信息可视化和探索性分析）
This example is completely imaginary.
A grocery store sells Feta-cheese for use in salads. It notices that it is losing money on this product and therefore decides to stop selling it. However, after having stopped selling this product, the grocery store notices that it now loses far more money than it did before. The reason for this is that the few customers who bought this product were wealthy and also bought many other products at the grocery store. When the store stopped selling the product, the customers went to other stores where they could buy it and, at the same time, purchased all other products they needed there.
The grocery store has modern cash-registers and records all purchases in a large database, so the information about the customers’ purchase patterns were, in principle, available before the store made the erroneous decision of stopping to sell Feta-cheese. An Information Visualization analyst studying the grocery store’s database, in the state that it was before the erroneous decision was made, should by performing an Exploratory Analysis, in principle, have been able to foresee this, without having been told anything about the data in advance.
2.2 Scientiﬁc Visualization and Conﬁrmatory Analysis （科学可视化和验证性分析）
This example is from the IEEE Visualization 2006 conference, where a visualization contest will take place. The subject of the contest is the TeraShake 2.1 earthquake simulation data set and the task is, through visualizations, to answer ﬁve questions, such as ”Which regions produce wave reﬂections?”. There is thus, in this case, a clear hypothesis that some of the regions will produce wave reﬂections and this hypothesis must either be rejected or conﬁrmed by ﬁnding these regions. Using the terminology of Keim, the task is therefore to perform a “Conﬁrmatory Analysis” of the data.
3 Common Visual Data Exploration Techniques （常见的视觉数据探索技术）
In the following is listed a few examples of the techniques that are used in Information Visualization to explore datasets that initially are completely unknown.
3.1 Icons and Glyphs (图标和字形)
Markers are featureless objects for representing positions in scatter plots, but multiple, clearly distinguishable markers are sometimes used in one and the same scatter plot to visualize important information. While markers usually only have positional attributes, there are objects that are designed to represent many pieces of information simultaneously. Such objects are called icons or glyphs ([3,4]) and have attributes, such as position, size, shape, color, and orientation, to which data is mapped. The hope is that users, by visual exploration, may be able to ﬁnd patterns among the objects.
3.2 Brushing & Linking （关联更新技术）
Brushing & Linking is a widely-used visual data exploration technique, with which multiple different visualizations of a dataset are viewed simultaneously. This technique is often used to combine the advantages of multiple forms of data visualization techniques, but can also be used to e.g. allow simultaneously brushing of markers in multiple scatter plots, organized in a scatter plot matrix. The advantage is that the interaction involved is both unambiguous and simple to perform and therefore seem familiar and natural to most human beings. In the beginning of a visual data exploration process, all markers have a common color. The process consist of
- Selecting one of multiple views.
- Brushing selected markers in this view with a color.
- Inspecting the other (linked) views to see the effect
3.3 The Grand Tour (总览)
In “The Grand Tour” technique [5–7] high-dimensional data are considered as forming a shape in a high-dimensional space. and the method therefore consists of moving a 2D or 3D “window” around this shape, so that a low-dimensional projection of it becomes visible. Signiﬁcant features of the high-dimensional shape may thereby become distinguishable, and, if one is lucky, one may even be able to guess what kind of high-dimensional shape the data form.
This technique is often combined with “Brushing & Linking”, so that subsets of the markers at any time may be brushed with a color to make them distinguishable from the other markers.
4 Software （软件）
There are also differences in the software that typically is used for Scientiﬁc Visualization and Information Visualization. Consider for example the Visualization Toolkit (VTK) , which is an open source visualization system that dates back to 1993. It was designed to be robust, understandable, extensible, modular and maintainable. For that purpose, it offers scientists a consistent object-oriented programming model that in 1993 was considered sophisticated, where visualizations are created by linking together a series of objects into a pipeline.
Seen from an Information Visualization perspective, it is clear problem that the visualization pipeline is designed to be “single-pass”, where only the resulting visualizations can be interacted with and where interaction usually is limited to rotating the visualization. Information Visualization techniques, such as Brushing and Linking require continuous interaction with the a database containing the original dataset. This requirement is not listed above, which makes VTK unsuitable for Information Visualization.
However,many Scientiﬁc Visualization analysts study “ Computational Steering,” with which feedback can be given to running simulations, based on visualizations of early results from the simulations. The requirements for these kinds of software therefore approach the requirements that Information Visualization scientists place on the software they use.
5 Conclusion （结论）
This paper presented Scientiﬁc Visualization and Information Visualization with emphasis on the latter. It describe a number of Information Visualization visual data exploration techniques and ﬁnally compared the requirements for the software used in both disciplines.
1. Card, S. K., Mackinlay, J. D., Shneiderman, B.: Readings in Information Visualiza- tion: Using Vision to Think Morgan Kaufmann (1999)
2. Keim, D. A., Ankerst, M.: Visual Data Mining and Exploration of Large Databases Tutorial, Conf. on Principles of Knowlege Discovery and Data Mining Freiburg, Germany (2001)
3. Carr, D. B., Nicholson, W. L.: Evaluation of Graphical Techniques for Data in dimensions 3 to 5: Scatterplot Matrix, Glyph, and Stereo Examples Proceedings of the Section on Statistical Computing (1985) 229–235
4. Pickett, R. M., Grinstein, G.: Iconographic Displays for Visualizing Multidimen- sional Data Proceedings of the IEEE Conference on Systems, (1988) 514–519
5. Asimov, D.: The grand tour: A tool for viewing multidimensional data SIAM. Jour- nal of Science and Statistical Computing 6 (1985) 128–143
6. Buja, A. and Asimov, D.: Grand tour methods: an outline. Computing Science and Statistics: Proceedings of the Seventeenth Symposium on the Interface (1985) 63–67
7. Wegman, E. J.: The grand tour in k-dimensions Computing Science and Statistics: Proceedings of the 22nd Symposium on the Interface (1991), 127–136
8. Schroeder, W.J., Martin, K., Loresen, W.: The visualization Toolkit: An Object Oriented Approach to Computer Graphics, Third Edition Kitware, Inc. (2004)