Search

During the past decade, we have witnessed an explosive growth in our capabilities to both generate and collect data. Various data mining techniques have been proposed and widely employed to discover valid, novel and potentially useful patterns in these data. Data mining involves the discovery of patterns, associations, changes, anomalies, and statistically significant structures and events in huge collections of data. One of the key success stories of data mining research and practice has been the development of efficient algorithms for discovering frequent itemsets – both sequential (Srikant & Agrawal, 1996) and non-sequential (Agrawal & Srikant, 1994). Generally speaking, these algorithms can extract co-occurrences of items (taking or not taking into account the ordering of items) in an efficient manner. Although the use of sets (or sequences) has effectively modeled many application domains, like market basket analysis, medical records, a lot of applications have emerged whose data models do not fit in the traditional concept of a set (or sequence), but require the deployment of richer abstractions, like graphs or trees. Such graphs or trees arise naturally in a number of different application domains including network intrusion, semantic Web, behavioral modeling, VLSI reverse engineering, link analysis and chemical compound classification. Thus, the need to extract complex tree-like or graphlike patterns in massive data collections, for instance, in bioinformatics, semistructured or Web databases, became a necessity. The class of exploratory mining tasks, which deal with discovering patterns in massive databases representing complex interactions among entities, is called Frequent Structure Mining (FSM) (Zaki, 2002). In this article we will highlight some strategic application domains where FSM can help provide significant results and subsequently we will survey the most important algorithms that have been proposed for mining graph-like and tree-like substructures in massive data collections.

This paper considers the problem of finding the most central nodes in neighborhoods of a given network with directed or undirected links taking into account only local information. An algorithm that calculates ranking, taking into account the n-hop neighborhood of each node is proposed. The method is compared to popular existing schemes for ranking, using Spearman's rank correlation coefficient and other metrics. An extension to a faster algorithm which reduces the size of the examined network is described as well

The task of querying wireless sensor network (WSN) to retrieve data of interest is very significant and a wealth of query types have been proposed in the context of WSNs. This article describes the d-hop k-data coverage query problem, which is a novel query type, aiming at extracting "feature(s) distribution maps" from WSNs. This problem generalizes earlier research problems, like top-k, skyband, and d-hop dominating sets. For this problem, we provide a fully distributed solution, the DaCoN protocol, that avoids constructing a "network spanner", since such a structure requires an expensive initialization procedure, misses the notion of neighborhoods, and most importantly, it creates hot-spots of communication, that shorten the network lifetime (i.e., some nodes deplete their energy very fast). We have developed a simulator to study the performance and behavior of the DaCoN protocol for various sensornet topologies and data distributions, and the obtained results attested the energy-efficiency and effectiveness of the protocol.

Metasearch engines are a significant part of the information retrieval process. Most of Web users use them directly or indirectly to access information from more than one data sources. The cornerstone of their technology is their rank aggregation method, which is the algorithm they use to classify the collected results. In this paper we present three new rank aggregation methods. At first, we propose a method that takes into consideration the regional data for the user and the pages and assigns scores according to a variety of user defined parameters. In the second expansion, not all component engines are treated equally. The user is free to define the importance of each engine by setting appropriate weights. The third algorithm is designed to classify pages having URLs that contain subdomains. The three presented methods are combined into a single, personalized scoring formula, the global KE. All algorithms have been implemented in QuadSearch, an experimental metasearch engine available at http://quadsearch.csd.auth.gr.

The World Wide Web, or simply Web, is a characteristic example of a social network (Newman, 2003; Wasserman & Faust, 1994). Other examples of social networks include the food web network, scientific collaboration networks, sexual relationships networks, metabolic networks, and air transportation networks. Socials networks are usually abstracted as graphs, comprised by vertices, edges (directed or not), and in some cases, with weights on these edges. Social network theory is concerned with properties related to connectivity (degree, structure, centrality), distances (diameter, shortest paths), “resilience” (geodesic edges or vertices, articulation vertices) of these graphs, models of network growth. Social networks have been studied long before the conception of the Web. Pioneering works for the characterization of the Web as a social network and for the study of its basic properties are due to the work of Barabasi and its colleagues (Albert, Jeong & Barabasi, 1999). Later, several studies investigated other aspects like its growth (Bianconi & Barabasi, 2001; Menczer, 2004; Pennock, Flake, Lawrence, Glover, & Giles, 2002; Watts & Strogatz, 1998), its “small-world” nature in that pages can reach other pages with only a small number of links, and its scale-free nature (Adamic & Huberman, 2000; Barabasi & Albert, 1999; Barabasi & Bonabeau, 2003) (i.e., a feature implying that it is dominated by a relatively small number of Web pages that are connected to many others; these pages are called hubs and have a seemingly unlimited number of hyperlinks). Thus, the distribution of Web page linkages follows a power law in that most nodes have just a few hyperlinks and some have a tremendous number of links In that sense, the system has no “scale” (see Figure 1).

During the past decade, the World Wide Web became the most popular network in the World. WWW grows with a very fast speed, thus the information that can be found through it is huge. In the early 90s, the first search engines for the WWW appeared. The user could give some keywords and the system returned a number of URLs (uniform resource locators) that contained the keywords. The order of the URLs in the return list was initially based on the number of the keyword occurrences in each URL. Some more sophisticated systems were taking into account the importance and the frequency of the keywords. As WWW was growing, a simple keyword search could match hundreds of thousands of pages. A human can only check the first twenty or even some more of the URLs that the search engine returns. Consequently, the ordering of the search results became very important. The most important URLs that are related with the search keywords should be ranked first. The link analysis ranking (LAR) is an objective way to sort search results. There are many advantages of the LAR over older methods. First of all the ranking is feasible without getting any feedback from the users. It is also not necessary to store the content of the URLs, but only the links. Another advantage is that it is difficult for the site developers to cheat by repeating keywords in the documents and moreover it may be pre-computed for all URLs. There are even more benefits using LAR to sort the search results that make it the best method used so far.

It is a great pleasure to welcome you to the 2008 IPS/UMS/WNGS/MUS workshop held in conjunction with the International Conference on Multimedia and Ubiquitous Engineering (MUE2008), Hanwha Resort Haeundae, Busan, Korea. The workshop covers topics on ubiquitous multimedia services and security in ubiquitous multimedia systems. All the submitted papers have undergone rigorous review process by the technical program committee members and external reviewers. We have selected 17 best papers for presentation and publication in the conference proceedings.

The deployment of wireless sensor networks in many application areas, e.g., aggregation services, requires self-organization of the network nodes into clusters. Quite a lot of node clustering techniques have appeared in the literature, and roughly fall into two families; those based on the construction of a dominating set and those which are based solely on energy considerations. The former family suffers from the fact that only a small subset of the network nodes are responsible for relaying the messages, and thus cause rapid consumption of the energy of these nodes. The later family uses the residual energy of each node in order to direct its decision about whether it will elect itself as a leader of a cluster or not. This family's methods ignore topological features of the nodes and are used in combination with the methods of the former family. We propose a novel distributed clustering protocol for wireless sensor networks, based on a novel metric for characterizing the importance of a node, w.r.t. its contribution in relaying messages. The protocol achieves small communication complexity and linear computation complexity. Experimental results for various sensor network topologies show that the protocol generates only a few clusters, guaranteeing a small number of message relays thus improving network lifetime

Discrete sequence modeling and prediction is an important goal and challenge for Web environments, both wired and wireless. Web clients’ datarequest forecasting and mobile location tracking in wireless cellular networks are characteristic application areas of sequence prediction in such environments. Accurate data-request prediction results in effective data prefetching, which combined with a caching mechanism can reduce userperceived latencies as well as server and network loads. Also, effective solutions to the mobility tracking and prediction problem can reduce the update and paging costs, freeing the network from excessive signaling traffic. Therefore, sequence prediction comprises a very important study and development area. This chapter presents information- theoretic techniques for discrete sequence prediction. It surveys, classifies, and compares the state-of-the-art solutions, suggesting routes for further research by discussing the critical issues and challenges of prediction in wired and wireless networks.

Energy conservation and access efficiency are two fundamental though competing goals in broadcast wireless networks. To tackle the energy penalty from sequential searching, the interleaving of index with data items has been proposed. Although, quite important contributions exist on providing broadcast indexes, they have one or more of the following problems. Firstly, all of them assume total ordering among broadcast data, and none considers the more general case of partial ordering. Secondly, they are balanced structure, which does not fit the "linear (one-dimensional) structure" of the wireless medium, in which imbalanced structures may offer significant advantages. Thirdly, they do not take into account the skewness in the access pattern, which prohibits larger performance gains to be reaped. Finally, they require all index items to be of equal size, which may not always give the optimal performance. To cope with all these problems, we introduce a new imbalanced tree-structured index. The new index is shown to be a generalization of two previously proposed high-performance indexes, and it introduces for the first time the problem of indexing partially ordered broadcast data. We present an experimental analysis of the proposed method, contrasting it with competing techniques. The analysis exhibits the efficiency of the proposed index in reducing the energy consumption without noticeably worsening the access latency

Broadcasting is an effective means for disseminating information in wireless ad hoc networks. In this paper we propose a novel distributed broadcasting protocol in wireless ad hoc networks, which is based on an highly efficient metric for characterizing the importance of a node, with respect to its contribution in covering the local neighborhood. The protocol is reliable and achieves small communication complexity with linear in the number of nodes computation complexity. Experimental results for a large variety of network topologies show that the proposed algorithm is capable of generating small connected dominating sets, which guarantee a relatively small number of rebroadcasts.

The Web has become the primary means for information dissemination. It is ideal for publishing data residing in a variety of repositories, such as databases. In such a multi-tier system (client - Web server - underlying database), where the Web page content is dynamically derived from the database (Web-powered database), cache management is very important in making efficient distribution of the published information. Issues related to cache management are the cache admission/replacement policy, the cache coherency and the prefetching, which acts complementary to caching. The present chapter discusses the issues, which make the Web cache management radically different than the cache management in databases or operating systems. We present a taxonomy and the main algorithms proposed for cache replacement and coherence maintenance. We present three families of predictive prefetching algorithms for the Web and characterize them as Markov predictors. Finally, we give examples of how some popular commercial products deal with the issues regarding the cache management for Web-powered databases.

During the past years, we have witnessed an explosive growth in our capabilities to both generate and collect data. Advances in scientific data collection, the computerization of many businesses, and the recording (logging) of clients’ accesses to networked resources have generated a vast amount of data. Various data mining techniques have been proposed and widely employed to discover valid, novel and potentially useful patterns in these data.

During the past decade, we have witnessed an explosive growth in our capabilities to both generate and collect data. Various data mining techniques have been proposed and widely employed to discover valid, novel and potentially useful patterns in these data. Data mining involves the discovery of patterns, associations, changes, anomalies, and statistically significant structures and events in huge collections of data. One of the key success stories of data mining research and practice has been the development of efficient algorithms for discovering frequent itemsets – both sequential (Srikant & Agrawal, 1996) and non-sequential (Agrawal & Srikant, 1994). Generally speaking, these algorithms can extract co-occurrences of items (taking or not taking into account the ordering of items) in an efficient manner. Although the use of sets (or sequences) has effectively modeled many application domains, like market basket analysis, medical records, a lot of applications have emerged whose data models do not fit in the traditional concept of a set (or sequence), but require the deployment of richer abstractions, like graphs or trees. Such graphs or trees arise naturally in a number of different application domains including network intrusion, semantic Web, behavioral modeling, VLSI reverse engineering, link analysis and chemical compound classification. Thus, the need to extract complex tree-like or graphlike patterns in massive data collections, for instance, in bioinformatics, semistructured or Web databases, became a necessity. The class of exploratory mining tasks, which deal with discovering patterns in massive databases representing complex interactions among entities, is called Frequent Structure Mining (FSM) (Zaki, 2002). In this article we will highlight some strategic application domains where FSM can help provide significant results and subsequently we will survey the most important algorithms that have been proposed for mining graph-like and tree-like substructures in massive data collections.

Discrete sequence modeling and prediction is a fundamental goal and a challenge for location-aware computing. Mobile client’s data request forecasting and location tracking in wireless cellular networks are characteristic application areas of sequence prediction in pervasive computing, where learning of sequential data could boost the underlying network’s performance. Approaches inspired from information theory comprise ideal solutions to the above problems, because several overheads in the mobile computing paradigm can be attributed to the randomness or uncertainty in a mobile client’s movement or data access. This article presents a new information-theoretic technique for discrete sequence prediction. It surveys the state-of-the-art solutions and provides a qualitative description of their strengths and weaknesses. Based on this analysis it proposes a new method, for which the preliminary experimental results exhibit its efficiency and robustness.

Web cache replacement algorithms have received a lot of attention during the past years. Though none of the proposed algorithms deals efficiently with all the particularities of the Web environment, namely, relatively weak temporal locality (due to filtering effects of caching hierarchies), heterogeneity in size and origin of request streams. In this paper, we present the CRF replacement policy, whose development is mainly motivated by two factors. The first is the filtering effects of Web caching hierarchies and the second is the intention of achieving a balance between hit and byte hit rates. CRF's decisions for replacement are based on a combination of the recency and frequency criteria in a way that requires no tunable parameters.

The recent developments in computer and communication technologies gave rise to Personal Communication Systems. Due to the nature of the PCS, the bandwidth allocation problem arises, which is based on the notion of bandwidth-on-demand. We deal with the problem of how to predict the position of a mobile client. We propose a new algorithm, called DCP, to discover user mobility patterns from collections of recorded mobile trajectories and use them for the prediction of movements and dynamic allocation of resources. The performance of the proposed algorithm is examined against two baseline algorithms. The simulation results illustrate that the proposed algorithm achieves recall that is comparable to that of the baseline algorithms and substantial improvement in precision. This improvement guarantees very good predictions for resource allocation with the advantage of very low resource consumption.

Semistructured data arise frequently in the Web or in data integration systems. Semistructured objects describing the same type of information have similar but not identical structure. Finding the common schema of a collection of semistructured objects is a very important task and due to the huge volume of such data encountered, data mining techniques have been employed. Maintenance of the discovered schema in case of updates, i.e., addition of new objects, is also a very important issue. In this paper, we study the problem of maintaining the discovered schema in the case of the addition of new objects. We use the notion of "negative borders" introduced in the context of mining association rules in order to efficiently find the new schema when objects are added to the database. We present experimental results that show the improved efficiency achieved by the proposed algorithm.

Improving the performance of the Web is a crucial requirement, since its popularity resulted in a large increase in the user perceived latency. In this paper, we describe a Web caching scheme that capitalizes on prefetching. Prefetching refers to the mechanism of deducing forthcoming page accesses of a client, based on access log information. Web log mining methods are exploited to provide effective prediction of Web-user accesses. The proposed scheme achieves a coordination between the two techniques (i.e., caching and prefetching). The prefetched documents are accommodated in a dedicated part of the cache, to avoid the drawback of incorrect replacement of requested documents. The requirements of the Web are taken into account, compared to the existing schemes for buffer management in database and operating systems. Experimental results indicate the superiority of the proposed method compared to the previous ones, in terms of improvement in cache performance.

flEPUHfH : AMflEAOKAAAIEPrEIA KAl flAPArorH kPAElOY ETHN EAAAAA H ckîqot) xuv cWTjvtKÛv a^ineXûvuv, 1700 cxxdpia, uetûvcxai avvcxùc,. Exqv cnitpàvcia auxrj napdfovxat cntxpancÇia oxavûXia, axatpî&cç nat ncptnou 5.000.000 hl xpaaioû. Av cÇaipcaouuc ta napa&oaiaxd xonixd npoiâvxa - iiaupoààtp vr\ , pcxaiva - xo unâXotno ctvai xpaaC Xeuxô xai oxo 905 cm xpanéÇio . H axoiaixf) TiapafuYn oto XuPLô» nou napa- HiiàÇci, a vx tnpoounc vc i xo iiioâ xrçç napafuïnç xpaoioû xai xo dXXo \iiaô uotpdÇovxai oi cunopot nov ov^ xc vxpûvouv xo npoiôv nai napouat dÇouv oxaotuoxnxa xai oi auvcxaiptax ixcç opiravûactç not» napouaidÇouv npoo6o., The Greekvineyard area is now reduced to 170.000 ha. It produces dessert grapes, currants, sultanas and ay early amount of about 5.000.000 hi of wines. Few of these are specific mavrodafni and retsina wines; most of them are white wines and 90% are tables wines. Peasant-produced wines recently decreased sharply down to less than half of the production only; wines produced by merchants and cooperatives share almost equally the larger half : they generally prepare blended rather than vintage wines., La surface du vignoble grec (170.000 ha) diminue. Sa production consiste en raisin de table, dont la part augmente, en raisins secs, durement concurrencés, et en raisins de cuve qui procurent désormais moins de 5 millions d'hectolitres : 90% de vin de table, 60% de blancs, quelques vins de cépage, de rares crus, un peu de vin doux naturel, etc., et productions spécifiques : mavrodafni et retsina. La production villageoise (moins de la moitié du total) s'est effondrée, mais des vins de propriétaires apparaissent; coopératives et marchands se partagent l'autre moitié et proposent en général des assemblages et, plus récemment, des vins de pays., Péchoux Pierre-Yves, Katsaros Dimitri. Vignes et vins en Grèce. In: Méditerranée, troisième série, tome 65, 3-1988. Vignobles et vins dans les pays méditerranéens, sous la direction de R. Courtot . pp. 45-54.

In a previous study, about 10 per cent of patients with regional enteritis were found to be deficient in glucose-6-phosphate dehydrogenase (G-6-PD). In the present study, enzymatic assay was performed in 69 patients with regional enteritis and 49 patients with ulcerative colitis. No examples of G-6-PD deficiency were found, so that it is concluded that G-6-PD deficiency is unlikely to be an etiologic factor in either of these diseases.