| 內容/備註 |
Abstract
The desire to augment our 3-dimensional perception and the need to understand multivariate problems spawned several multidimensional visualization methodologies. Starting from early successes of visualization, like Dr. J. Snow¡¦s dot map in 1854 showing the connection of cholera to a water pump in London, many techniques have been developed. Some will be reviewed so as to establish the connection between multivariate problems a multidimensional geometry. Understanding the underlying geometry of a multivariate problem provides important insights into what is possible and what is not. For the unambiguous visualization of multidimensional geometry and, in turn, multivariate relations Parallel Coordinates ¡V the leading Multidimensional Visualization Methodology ¡V is introduced and rigorously developed. Relations among N real variables are mapped uniquely into subsets of 2-space having geometrical properties enabling the visualization of the corresponding N-dimensional hypersurfaces. After the basic representation results, associated algorithms for constructions, intersections, transformations, containment queries, proximity and others will be presented. The development is interlaced with applications of the relevant results starting with demonstrations of Data Mining on real datasets (i.e. Fea- ture extraction from LandSat data, Financial, Process Control, Pilot Selection, Raising the Yield and Quality of VLSI chips, and others). They are followed by Collision Avoidance Algorithms for Air Traf- fic Control which are based on the representation of lines in multidimensional space. The detection of coplanar points and the representation of planes and hyperplanes lead to some applications in Computer Vision, Geometric Modeling and elsewhere. More examples of Visual Data Mining are given. An efficient geometric automatic classifier algorithm is motivated and is demonstrated on some challenging datasets. Finally, the representation of curves and hypersurfaces is taken up together with interactive applications to Process Control, Instrumentation and Heuristic Optimization. Nonlinear VISUAL models, in terms of hypersurfaces, are constructed from data and used interactively for Decision Support, Sensitivity Analysis, studying feasibility and effect of constraints as well as trade-off analysis.
NOTE: Do not be intimidated by the formalistic language. The organizer is also well known for numerological anecdotes and palindromic diversions. Valuable prizes will be distributed in real-time to those contributing memorable and noteworthy digressions.
KEYWORDS: Multidimensional Geometry, Multidimensional/Multivariate Visualization, Information Visualization, Parallel Coordinates, Visual & Automatic Data Mining, Intelligent Process
Control & Instrumentation, Nonlinear Modeling, Decision Support.
※全程以英文上課共計21小時
對象：全國大學、技職院校、廠商對顯示科技的解決方法論有興趣，並願意投入此方面教學及研究的學生(undergraduate, postgraduate students) 、教師與廠商
報名方式：請事先報名www.eecs.stut.edu.tw/→點選 研討會議
報名費用：校外人士收費1000元
主辦單位：成功大學電機系,南台科技大學電子系, 中央研究統計所,南台科技大學育成中心,教育部光機電成大中心
※ 若有任何問題請洽賴培淋 老師 0935031061
|