The Experts below are selected from a list of 9 Experts worldwide ranked by ideXlab platform
Auer Michael - One of the best experts on this subject based on the ideXlab platform.
-
User generated Cartography by crowd sourcing web map styling
Digital Proceedings of the 25th International Cartographic Conference (ICC), 2011Co-Authors: Auer MichaelAbstract:BACKGROUND AND OBJECTIVES For a long time the production of maps has been the domain of skilled specialists who had profound knowledge of the theoretical and practical basics of Cartography. This is changing with the advances of communication and web technologies. The user himself has more and more influence on what and how things are displayed on the maps he uses. He can adapt the maps to his own needs and on the other hand he is also able to create maps for other people to share geographical information. But how does he know about the proper design? Cartography has gone many steps, developing from a craft with artistic features to a discipline, which tries to scientifically determine the concepts of how maps work. Within all those steps, the user has been focused more and more, as the cartographer tries to make maps, which are functional, usable and understandable through the eyes of others. So the main issue in making maps for others is to know about the users. Their context, their abilities, their experiences and expectations on what is seen on a map. Cartography and the mapmaking user have to find out about the cognitive schemata, which the map readers have in their minds, to be able to adapt the map representation to the mental representations of the readers. The assumption made here is that a better map design is reached by minimizing the distance or the differences between the physical and the mental representation. Today many successful Examples of Web2.0-Services show that new possibilities of communication provide new ways of gathering individual knowledge together to a greater whole and use it for the benefit of all. Those new possibilities of communication implicate as well new possibilities for cartographic research and map production. Several well-known VGI-Projects (see Goodchild 2007) like Wikimapia or OpenStreetMap have demonstrated how powerful a simple data collection platform can be used to crowd-source the creation of worldwide geodatasets. Although there are many ways the Web2.0 can be used to create and publish cartographic works, it is not the Web2.0 itself, which guarantees good quality. Cartwright 2008 shows some poor examples of Mashup maps, where default symbols and basemaps from a web map API were used in mobile map contexts, ignoring map extent and resulting in poorly placed and overlapping marker icons. So to take advantage of the Web2.0 for map quality issues will depend on how the users can interact with the information provided by other users. It " s the collaborative manner in which the Web2.0 has to be applied to benefit from the crowd intelligence. It even Doesn " t have to be a synchronous process, it can be an asynchronous collaboration like, for example in Wiki projects. The possibility of cross checking and giving feedback about the quality of data and the quality of design, as well as the merging or puzzling together of information from different users are the strengths of the collaborative crowd-sourcing approach. Now the aim of this paper is to point out, that and how we can use principles of Web2.0-communication (see O " Reilly 2005) in a collaborative way to create " good " map styles. A " good " map style, in this sense, should first contain the content a user expects in a certain context and second the presentation of this content should match the viewing habits of the reader. In other words, a " good " map style should result in a map, which matches closely to the specific map schemata (see MacEachren 1995) a user has developed before, in the given context or at least, which matches closely to a general map schema which the reader can apply, while reading the map. The aim here is to minimize the mental costs in retrieving information from the map. Taking this into account, the map design should depend on the map purpose or function and the characteristics of the map users. Both will influence the map design, the purpose which defines the context of usage (Functionality, Actions, Media etc.) and the users with their abilities of visual perception and cognitive models (disabilities, language, culture, map reading skills etc.) , which define the expected content and its appearance. To create maps for more than one person always means to encounter several individual and therefore different mental concepts. This implicates that the map style has to mediate between the different mental user requirements. The following chapter discusses the possibility to organize the process of gaining insights into the user expectations, with respect to content and appearance by taking the advantages of the crowd-sourcing principle, giving the users themselves the chance to state their individual conception of the " correct " selection and representation of geoobjects and further derive a combined community-based map-conception, which will be a prototypical map style in relation to the contributing community members. The results chapter outlines a first attempt of realizing an online map style editor called SLDExplorer to create or modify StyledLayerDescriptor Documents, which is based on open standards of the Open Geospatial Consortium. Finally the paper closes with some remarks about the future work towards a crowd-sourcing solution of cartographic web map styling. APPROACH AND METHODS For such a process of data acquisition of course there must be a certain infrastructure which enables the user to try out and develop design preferences, there must be a format or a formal language which allows to express individual conceptions of representations and there must be a platform where you can save your concept and share it with others. There is an existing example of a Web2.0 web map style editor provided by Cloudmade, which implements a good part of the infrastructure and enables users to interactively create their individual OSM-Style by choosing colors, outline appearance and selecting which objects should appear at which zoom levels. The styles can be shared with the public to be used in web map Applications over the cloudmade API and can also be chosen as a base to create new user styles. It " s free of charge and it hosts the rendered map tiles, so that the user Doesn " t have to care about server infrastructure and tile caching technology etc. But at a closer look it " s obvious, that this Web2.0 Application Doesn " t make use of the crowd intelligence like Wikipedia or OpenStreetMap. What is missing is the collaborative approach on working together on the same content in order to raise quality. In fact all styles are made by single users and don " t get better if more users are participating creating more styles. Also the authors can " t state the purpose for which the maps are made for, which makes it hard to select a style and evaluate its quality for a use in another user " s context. The following conceptual approach takes those disadvantages into account and outlines a more generic way, which is not bound to specific data. CONCEPTUAL APPROACH Figure 1 depicts a conception of a crowd-sourcing-process for interactive web map styling. The principle which takes effect here is based upon the fact that the individual user reveals his specialized knowledge about his concept of desired or necessary content, plus his preferences in relation to the used map symbols for the visual representation. The collection of all the individual information in a database makes it possible to determine on the one hand a prototypical content and on the other hand a prototypical style (Sign-Object-Reference) of a map, which is adapted to the user-community and the map purpose. There should also be several feedback loops of already stored information from other users which allow the user to get some decision help and orientation at the creation of their individual object selection and sign-object-references. This way it is possible to learn from other users who made decisions in the same situation before. In case that the user hasn " t developed a specific map schema yet, he can adopt the schemas from other community members and train his own map schema with the side effect of converging to a common map schema. The crowd-sourcing-process starts with the formulation of the idea of the map to be produced (1). As the design of a map depends among other things of the map purpose which the map shall fulfill and of the target-group which will use and has to understand the map, at first the purpose has to be described carefully as this will be given to the users as a guiding idea during the process enabling them to imagine the usage context. Maybe there is also a proposal of an appropriate dataset to use which can be added to the idea but this could also be a point of community discussion to bring other or better available datasets into play. Once finished with the description of the idea the crowd-sourcing-process can go on by the act of communicating the idea to the community or the target-user-group (2). Now this group can design a common map (3) in several steps by each other revealing their expectations of the required content and checking if those expectations can be fulfilled by the proposed dataset. Out of the expected and available contents a list of every user develops with each ones individual object selections from which then by means of statistical analysis a common prototypical set of expected and available objects can be derived. This step closes a kind of user analysis which gives us information about important expected contents of the map for the achievement of the given map purpose. Fig. 1: Process of crowd-sourcing interactive web map styling
Josep Tarrés I Turon - One of the best experts on this subject based on the ideXlab platform.
-
Project management: a preliminary graphical proposal
2010Co-Authors: Josep Tarrés I TuronAbstract:Project management uses tools to program (plan and implement) projects. One of them is Microsoft Project, a powerful computer program with concepts, structure and terminology that are sometimes confusing, inappropriate or incomplete. Although the programmers took care of these matters, the Application structure is hard to understand and follow. All of this became apparent when the author taught MS Project during his twelve years of teaching at university and as a professional consultant. This paper shows the interrelation of the different project management variables (including quality, time, and cost) and the graphical representation of the project allowing the creation of a graphical proposal that not only maintains coherence in the use of the symbols of variables or interrelations but also provides what the Application Doesn�t. This proposal has completed its conception phase and is adequate for a graphical interface that is at stage of development.
Qingren Wang - One of the best experts on this subject based on the ideXlab platform.
-
Research on the consistency of LVQ classifier
International Conference on Image Processing and Pattern Recognition in Industrial Engineering, 2010Co-Authors: Qing-wen Zhou, Kai Wang, Qingren WangAbstract:ABSTRACT As a self-organizing artificial neural network model based on supervised learning, the LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. The improved LVQ classifier have been greatly developed in previous works, and the experimental results on specific problems show that the improved LVQ classifier is indeed better than the standard learning algorithms proposed by Kohonen. Different from previous works, the consistency is studied in this paper to provide a theoretical support for the LVQ classifier. Furthermore, a simulation is included in this paper to provide an experimental support for our theoretical work. Keywords: consistency; learning vector quantization; pattern recognition 1. INTRODUCTION The Learning Vector Quantization (LVQ) classifier is a self-organizing artificial neural network model based on supervised learning proposed by Kohonen. The LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. In the practice of the LVQ, two main problems have been found in Kohonens standard learning algorithm [2~4] due to its hard competitive learning strategy, which are neurons underutilization and information waste between input sample and competitive neuron. In view of these problems, many improved LVQ classifiers have been proposed based on fuzzy technique [5~6]. On the other hand, with the development of statistical learning theory and the Support Vector Machine, researchers recognize the advantages of interval maximizing in the design of classifiers, which further improves the efficiency of LVQ classifier [7~8]. In particular, the improved LVQ classifier proposed in [8] achieves similar accuracy with SVM but faster classification speed in the research on the text classification. Other research on the LVQ classifier improvement are not listed here, such as how to initialize LVQ, etc. The main purpose of the research on th e LVQ above is to improve the performance of solving practical problems. In this article, we mainly focus on the consistency of the LVQ classifier in order to provide theoretical support to the development of the LVQ classifier. The consistency of the LV Q classifier means whether LVQ classifier can achieve the Bayesian optimal in the case of an infinite sample space (ideal case). Consider the particularity of the research target, the experiment was not based on a specific Application. (This is ma inly due to that the Bayesian optimal recognition rate of a practical Application cant be estimated, so it is difficult to validate the theoretical result. On the other hand, the researc h target of this paper is not to improve the performance of LVQ and also didnt propose an improved algorithm. Therefore, the experiment on a specific Application Doesnt make sense. ) So we only provide experimental support for theoretical result through the experiment on artificial simulation data set.
Qing-wen Zhou - One of the best experts on this subject based on the ideXlab platform.
-
Research on the consistency of LVQ classifier
International Conference on Image Processing and Pattern Recognition in Industrial Engineering, 2010Co-Authors: Qing-wen Zhou, Kai Wang, Qingren WangAbstract:ABSTRACT As a self-organizing artificial neural network model based on supervised learning, the LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. The improved LVQ classifier have been greatly developed in previous works, and the experimental results on specific problems show that the improved LVQ classifier is indeed better than the standard learning algorithms proposed by Kohonen. Different from previous works, the consistency is studied in this paper to provide a theoretical support for the LVQ classifier. Furthermore, a simulation is included in this paper to provide an experimental support for our theoretical work. Keywords: consistency; learning vector quantization; pattern recognition 1. INTRODUCTION The Learning Vector Quantization (LVQ) classifier is a self-organizing artificial neural network model based on supervised learning proposed by Kohonen. The LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. In the practice of the LVQ, two main problems have been found in Kohonens standard learning algorithm [2~4] due to its hard competitive learning strategy, which are neurons underutilization and information waste between input sample and competitive neuron. In view of these problems, many improved LVQ classifiers have been proposed based on fuzzy technique [5~6]. On the other hand, with the development of statistical learning theory and the Support Vector Machine, researchers recognize the advantages of interval maximizing in the design of classifiers, which further improves the efficiency of LVQ classifier [7~8]. In particular, the improved LVQ classifier proposed in [8] achieves similar accuracy with SVM but faster classification speed in the research on the text classification. Other research on the LVQ classifier improvement are not listed here, such as how to initialize LVQ, etc. The main purpose of the research on th e LVQ above is to improve the performance of solving practical problems. In this article, we mainly focus on the consistency of the LVQ classifier in order to provide theoretical support to the development of the LVQ classifier. The consistency of the LV Q classifier means whether LVQ classifier can achieve the Bayesian optimal in the case of an infinite sample space (ideal case). Consider the particularity of the research target, the experiment was not based on a specific Application. (This is ma inly due to that the Bayesian optimal recognition rate of a practical Application cant be estimated, so it is difficult to validate the theoretical result. On the other hand, the researc h target of this paper is not to improve the performance of LVQ and also didnt propose an improved algorithm. Therefore, the experiment on a specific Application Doesnt make sense. ) So we only provide experimental support for theoretical result through the experiment on artificial simulation data set.
Kai Wang - One of the best experts on this subject based on the ideXlab platform.
-
Research on the consistency of LVQ classifier
International Conference on Image Processing and Pattern Recognition in Industrial Engineering, 2010Co-Authors: Qing-wen Zhou, Kai Wang, Qingren WangAbstract:ABSTRACT As a self-organizing artificial neural network model based on supervised learning, the LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. The improved LVQ classifier have been greatly developed in previous works, and the experimental results on specific problems show that the improved LVQ classifier is indeed better than the standard learning algorithms proposed by Kohonen. Different from previous works, the consistency is studied in this paper to provide a theoretical support for the LVQ classifier. Furthermore, a simulation is included in this paper to provide an experimental support for our theoretical work. Keywords: consistency; learning vector quantization; pattern recognition 1. INTRODUCTION The Learning Vector Quantization (LVQ) classifier is a self-organizing artificial neural network model based on supervised learning proposed by Kohonen. The LVQ classifier has been widely applied and deeply studied due to its good practical performance on the pattern recognition problems. In the practice of the LVQ, two main problems have been found in Kohonens standard learning algorithm [2~4] due to its hard competitive learning strategy, which are neurons underutilization and information waste between input sample and competitive neuron. In view of these problems, many improved LVQ classifiers have been proposed based on fuzzy technique [5~6]. On the other hand, with the development of statistical learning theory and the Support Vector Machine, researchers recognize the advantages of interval maximizing in the design of classifiers, which further improves the efficiency of LVQ classifier [7~8]. In particular, the improved LVQ classifier proposed in [8] achieves similar accuracy with SVM but faster classification speed in the research on the text classification. Other research on the LVQ classifier improvement are not listed here, such as how to initialize LVQ, etc. The main purpose of the research on th e LVQ above is to improve the performance of solving practical problems. In this article, we mainly focus on the consistency of the LVQ classifier in order to provide theoretical support to the development of the LVQ classifier. The consistency of the LV Q classifier means whether LVQ classifier can achieve the Bayesian optimal in the case of an infinite sample space (ideal case). Consider the particularity of the research target, the experiment was not based on a specific Application. (This is ma inly due to that the Bayesian optimal recognition rate of a practical Application cant be estimated, so it is difficult to validate the theoretical result. On the other hand, the researc h target of this paper is not to improve the performance of LVQ and also didnt propose an improved algorithm. Therefore, the experiment on a specific Application Doesnt make sense. ) So we only provide experimental support for theoretical result through the experiment on artificial simulation data set.
