Abstract In few-shot learning, the model must first be trained using the same type of dataset, and its effectiveness may be compromised when identifying classes from different datasets. Therefore, recent research has proposed cross-domain few-shot learning, which aims to train a single-network model that can be applied across most datasets. Cross-domain few-shot learning typically involves two stages during training: (1) pre-training and (2) meta-training. In the pre-training stage, a feature extractor is trained using traditional supervised learning to extract features. In the meta-training stage, a classifier is trained using few-shot learning methods, whereby meta-learning algorithms classify the features extracted by the feature extractor. In this study, a dual attention mechanism that incorporates channel and spatial attention is introduced into the feature extractor in the pre-training stage, to enhance feature extraction capabilities. In the meta-training stage, samples obtained from Fourier-transform-based image corruption algorithms are introduced. The transformation algorithm scales the phase and amplitude of the Fourier transform output to change the texture and semantics of the image, allowing the classifier to learn more diverse images and improve its generalization ability. Finally, during the meta-training process, the prediction losses of the images before and after corruption are simultaneously calculated to maintain consistency in the prediction results and effectively mitigating the risk of overfitting. The experimental results demonstrate that the proposed method achieves better performance on the majority of the eight testing datasets, effectively reducing overfitting during training.

Abstract Since futuristic cellular core networks are envisioned to accommodate emerging vertical domain applications with diverse QoS requirements, software‐defined networking (SDN) technology has played a vital role in improving the scalability and flexibility of network management. This paper presents the design and development of a service‐aware transport layer framework in SDN‐enabled cellular core networks. It aims to study and accommodate three typical traffic types: enhanced mobile broadband (eMBB), ultra‐reliable and low latency communications (URLLC), and massive machine‐type communications (mMTC) in optimal path selection. The proposed framework is able to dynamically allocate network resources to corresponding traffic of above types for satisfying the requirements of each identified flow, using a multi‐level queue management mechanism to alleviate the flow dropping problem when the network becomes congested. For experiment and verification, a testbed platform was built to evaluate the functionalities of developed modules, the performance of the diffserv‐based optimal path selection, and the operation of multi‐level queue management. The experiment results demonstrate that the proposed algorithms could obtain a higher accepting ratio than the other ones in different network conditions.

Abstract This paper proposes an architecture based on the improved local binary pattern (LBP) shallow deep convolution neural network, which integrates hand‐crafted feature pre‐processing and the advantage of character learning in the supervised high‐level function of CNN, in order to enhance its performance. This study introduced the information of scale space into the LBP to reduce the sensitivity to noise, and applied feature maps with two features, the maximum selection feature map (MLBP) and the first selection feature map (FLBP). The former selected the edge with the strongest intensity to reduce the influence of noise points, while the latter measured local binary features through the scale detection of an effective edge. In the network architecture design, according to the differences of input features, networks of different depths were used for learning, and the features learned by the two networks were adopted for classification. The experimental results show that, the ILBPSDNet proposed had certain recognition abilities in many character data sets, and the network parameters and computation were also reduced. Therefore, it has a significant effect in realizing the application of real‐time character recognition. Finally, compared with other latest networks, its network performance could be maintained at a certain level.

Data clustering is a well-known data analysis methodology in the optimization community that can be formulated as a Np-hard problem. The main concept of the clustering procedure is to identify and assign data objects into sensible disjointed groups based on similarity measures, such as the Euclidean distance. In this case, data objects within a group are similar to each other with regard to the similarity measure, and each group is dissimilar to each other. Fuzzy clustering is also a popular and powerful research method with many applications, such as medical imaging processing with real words. In the fuzzy clustering research field, the fuzzy c-means (FCM) algorithm is one of the most important fuzzy clustering approaches for efficient and easy implementation; however, the FCM algorithm has a weakness in which it can easily become trapped in local optima. Thus, some metaheuristic algorithms, such as genetic algorithm (GA) and particle swarm optimization (PSO), can be combined with FCM to solve the clustering problem. Recently, a novel metaheuristic algorithm inspired by the social behavior of humpback whales, called the whale optimization algorithm (WOA), has been proposed. In addition, the WOA has obtained more efficient solutions than some popular metaheuristic algorithms, such as the gene GA and PSO. In this paper, a fuzzy clustering algorithm that called the memetic fuzzy whale optimization (MFWO) algorithm, is proposed to solve the data clustering problem. The performance of the proposed algorithm is evaluated using eight real-world UCI benchmarks, and the FCM and fuzzy PSO algorithms are compared. The experimental results show that the proposed MFWO algorithm is more efficient than the compared algorithms.

The vehicle routing problem is a critical research issue in logistics because the strategies used may strongly impact its performance. A good strategy would allow the management to make a good decision to enhance the performance of logistics. In practice, we have to consider two or more factors in solving the vehicle routing problem at the same time. An effective hybrid search algorithm for the vehicle routing problem with time windows is presented in this study, which takes into account at the same time the number of vehicles and the total distance for vehicles based on a novel metaheuristic algorithm named coral reefs optimization algorithm with substrate layers (CRO-SL). The experimental results show that the proposed algorithm outperforms all the other state-of-the-art algorithms compared in this paper.

Since the number of vehicles on the road has been dramatically increased, how to find out a “good solution” to improve the traffic situation is a difficult optimization problem even if there is still plenty of room for improvement. The main focus of this study is on improving the quality of solutions for the traffic light optimization problem to minimize the waiting time of all the vehicles and maximize the number of vehicles arriving at the destination within a certain time period. An effective hybrid metaheuristic algorithm, called grey wolf grasshopper hybrid algorithm (GWGHA), that leverages the strength of grey wolf optimizer and grasshopper optimization algorithm will be presented in this paper to find a better result in solving the cycle traffic light problem (CTLP). The solutions obtained are simulated with a well-known microscopic traffic simulator named Simulation of Urban MObility (SUMO). Experiments were carried out based on data from Kaohsiung, Taiwan, Bahıa Blanca, Argentina, and Ma laga, Spain that represent different experimental cases to evaluate the performance of GWGHA. The simulation results show that the proposed algorithm is superior to the state-of-the-art search algorithms compared in this paper for solving the CTLP.

An improved quantum-inspired evolutionary algorithm (iQEA) is presented in this paper to improve the clustering result of a data clustering problem. Like the other QEA-based algorithms, the iQEA uses Q-bits to denote the state of a quantum particle and Q-gate as an evolutionary operator to guide the search directions. Unlike the fixed rotation degree of QEAs, the rotation degree of iQEA will be changed at different iterations. Experimental results show that the iQEA is able to find a better result than all the other metaheuristic algorithms compared in this paper in terms of quality.

How to reduce the computation time and how to improve the quality of the clustering result are the two major research issues. Although several efficient and effective clustering algorithms have been presented, none of which is perfect. As such, an effective clustering algorithm, which is based on the prediction of searching information to determine the search directions at later iterations and employs the k-means as the local search operator to fine-tune the end result, is presented in this paper. Simulation results show that the proposed algorithm is less sensitive to the initial random solution; thus, it is capable of providing a better result than the other clustering algorithms compared in this paper in terms of the quality of the clustering result.

One of the most well-known clustering methods for wireless sensor network is, no doubt, the so-called low energy adaptive clustering hierarchy (LEACH) because it is simple and easy to implement. Although LEACH tries to provide a fair selection mechanism by randomly selecting a number of sensors as the cluster-heads, it does not take into account the distribution of sensors, the main reason that LEACH is not able to allot transmissions to sensors efficiently, so it will make sensors far from the base station consume more energy in some cases. An efficient clustering algorithm to overcome this problem is presented in this paper. The proposed algorithm leverages the strength of a novel metaheuristics, search economics, and LEACH-centralized (LEACH-C) for wireless sensor network (WSN). Simulation results show that the search economics based clustering algorithm is capable of not only prolonging the lifetime of a WSN but also providing a balance strategy for the energy consumption of sensors in a homogeneous WSN.

A human action recognition system based on image depth is proposed in this paper. Depth information features are not easily disturbed by noise; and due to this characteristic, the system can quickly extract foreground targets. Moreover, the target data, namely, depth and two-dimensional (2D) data, are projected to three orthogonal planes. In this manner, the action featured in the depth motion along the optical axis can clearly describe the trajectory. Based on the change of motion energy and the angle variations of motion orientations, the temporal segmentation (TS) method automatically segments the complex action into several simple movements. Three-dimensional (3D) data is further applied to acquire the three-viewpoint (3V) motion history trajectory, whereby a target’s motion is described through the motion history images (MHIs) from the 3Vs. The weightings corresponding to the gradients of the MHIs are included for determining the viewpoint that bests describe the target’s motion. In terms of feature extraction, the application of multi-resolution motion history histograms can effectively reduce the computational load and achieve a high recognition rate. Experimental results demonstrate that the proposed method can effectively solve the self-occlusion problem.

The clustering problem is among the most important optimization problems. Given that it is an NP-hard problem, it can be efficiently solved using meta-heuristic algorithms such as the gravitation search algorithm (GSA). GSA is a new swarm-based algorithm particularly suitable for solving NP-hard combinatorial optimization problems. This paper solves the clustering problem with a newly proposed memetic GSA (MGSA) algorithm. MGSA is coupled with the pattern reduction operator and the multi-start operator. The proposed MGSA algorithm was verified on six UCI benchmarks and images segmentation. Based on a performance comparison amongst MGSA, the original GSA, and two state-of-the-art meta-heuristic algorithms (Firefly algorithm and the Artificial bee colony algorithm), we observe that the proposed algorithm can significantly reduce computation time without compromising much on the quality of the solution.

This research examines an assessment system that combines a mobile device with NFC (Near Field Communication) technology. Teachers attach or embed NFC tags into answer choices and students answer each question by bringing their mobile phone close to the answer that they think is the correct one. A four-week experiment shows that this assessment method takes more time than a traditional one. Also, the grades differ from those using traditional assessment. Analysis of the experiment results and subsequent tester survey and interviews show that this assessment method can help students incorporate what they have learned in the classroom into real life, as well as reflects how well students comprehend what is taught to them in a classroom setting.

Rule-based scheduling algorithms have been widely used on many cloud computing systems because they are simple and easy to implement. However, there is plenty of room to improve the performance of these algorithms, especially by using heuristic scheduling. As such, this paper presents a novel heuristic scheduling algorithm, called hyper-heuristic scheduling algorithm (HHSA), to find better scheduling solutions for cloud computing systems. The diversity detection and improvement detection operators are employed by the proposed algorithm to dynamically determine which low-level heuristic is to be used in finding better candidate solutions. To evaluate the performance of the proposed method, this study compares the proposed method with several state-of-the-art scheduling algorithms, by having all of them implemented on CloudSim (a simulator) and Hadoop (a real system). The results show that HHSA can significantly reduce the makespan of task scheduling compared with the other scheduling algorithms evaluated in this paper, on both CloudSim and Hadoop.

Designing a high-quality assessment system is a challenge for college-level educators. An effective assessment system requires mastering the professional knowledge and skills involved. The field of assessment and evaluation has developed many important concepts, principles, and methods to guide practice. This study proposes an intelligent assessment management system for the summative assessment of college courses. It integrates a virtual online assessment management system and in-classroom examinations. Before an exam, the scope of the examination is selected by teachers, and students who will take the exam are listed. According to the number of students, the system configures the range of examination rooms. Students can go to any of the examination rooms to take the exam during the exam period. Each table has a near-field communication (NFC)-enabled tablet for student identification via a student APP. The identification process uses digital certificates to ensure uniqueness and to avoid substitute testers. Invigilators use an exam management APP to access information about the students taking the exam in the exam room, and give feedback regarding the real-time situation to the system. Instructors use the assessment management APP to check students' situation after the exam. The system ensures that all students receive fair treatment during the exam. The proposed system improves the effectiveness of assessment.

There is no doubt about the potentials of smart grid, for the traditional power grid has been kind of out of date, in terms of not only its infrastructure but also its restrictions on the way information is communicated. To provide better and smarter services via smart grid, the first thing we have to do is to make it more intelligent. Among the technologies that can be applied to smart grid to make it more intelligent, data mining will certainly play a vital role. This paper begins with a brief introduction to smart grid, followed by a discussion on the supervised learning (classification) and the unsupervised learning (clustering). Several open and possible research issues are then given to depict the future trends of smart grid.

ABSTRACT As the concept of radio‐frequency identification (RFID) relay attack has been successfully implemented and demonstrated, the research of RFID distance‐bounding protocols to deter RFID relay attacks has drawn much attention from both the industry and academia. Conventionally, researchers adopted linear composition of secrets to resist terrorist fraud attacks. Recently, Peris‐Lopez et al. studied the weaknesses of previous RFID distance‐bounding protocols and proposed that non‐linear composition of secrets and inclusion of more random nonce could help RFID resist key disclosure attack and terrorist fraud attack. In this paper, we will show that non‐linear composition of secrets cannot help enhance the security actually. Copyright © 2012 John Wiley & Sons, Ltd.

Since the high computational cost of the structure-based protein-Ligand docking prediction is one of the major problems in designing new drugs, many researchers keep looking for a high performance search algorithm to find the workable directions to drug design as well as a simulator platform being able to test and verify the new drugs. In this paper, an improved version of Lamarckian genetic algorithm (ILGA) is first presented for enhancing the performance of LGA by using pattern reduction to reduce the computation cost and using tabu search to increase the search diversity to further find the better results. In addition, the proposed algorithm is also applied to a well-known simulator platform (AutoDock) to evaluate the performance of the proposed algorithm. The simulation results show that the proposed algorithm can enhance the performance of ILGA in terms of convergence performance especially for highly flexible ligands.

This paper presents a novel pheromone update strategy for improving the clustering results of ant colony optimization (ACO). The proposed algorithm is motivated by the observation that most of the ACOs only keep track of the promising foraging information, which has the potential to lead to better solutions than all the other search directions in the pheromone table. This eventually makes the search converge to particular search directions in later iterations because the pheromone values on good routing paths will be reinforced. As such, the breadth of search (diversity) will be reduced, thus limiting the clustering results of ACO. The proposed algorithm adds a second pheromone table to ACO for recording the unpromising foraging information that is worse than all the other search directions and using a novel construction method to explore the new search directions. In other words, by leveraging the strengths of diversification and intensification, the proposed algorithm can find better solutions than traditional ACO. To evaluate the performance of the proposed algorithm, we use it to solve the data clustering problem. Our experimental results indicate that the proposed algorithm can significantly improve the quality of ant colony optimization.

This paper presents a novel ant colony optimization (ACO) based on an efficient solution construction strategy (transition operator) for improving the quality of the end results of job shop scheduling problem (JSSP). Inspired by the observation that the quality of the end results of ACO is largely affected by their operators-especially the transition operator, a novel solution construction strategy is presented in this paper. The proposed algorithm uses two different strategies to compute the probability of solution construction to improve the end results. Our experimental results show that the proposed algorithm outperforms all state-of-the-art job shop scheduling algorithms evaluated in this paper and can significantly improve the quality of ant colony optimization for JSSP.

This paper presents a novel algorithm for tagging the relationship of test on the learning content called Binary Union-plus Tag (BUT). The proposed algorithm is motivated by the desire of students to re-study unacquainted content after a test. By binary coding items of content, BUT provides the users of an e-Learning system a certain range of content which lacks learning to connect to tests and content. It starts off with one of the content structures as a reference structure. Then, all the mistakes made on the test are mapped to the reference structure to highlight the re-studying area in the learning materials. To evaluate the performance of the proposed method, it has been used on the learning process of two hundred grade students. Moreover, the experimental results show that the proposed algorithm can considerably improve the achievement of students in terms of both locating the essential range of the content and providing the necessary matters.

By and large, population-based metaheuristics provide a better solution to the codebook generation problem of vector quantization (VQ) than Generalized Lloyd Algorithm (GLA) and single-solution-based metaheuristics. They are, however, all much slower. In this paper, we present an efficient method to speed up the performance of particle swarm optimization (PSO), called Fuzzy Pattern Reduction Enhanced Particle Swarm Optimization (FPREPSO). The proposed method first uses PSO to search for the global solutions. Then, it relies on pattern reduction to eliminate computations that are essentially redundant in the convergence process of PSO. And finally, it employs a set of fuzzy inference rules to decrease the chance of eliminating patterns that should not be eliminated. To evaluate the performance of the proposed algorithm, we compare it with GLA and GLA-based algorithms such as standard GLA, pattern reduction enhanced GLA, genetic k-means algorithm (GKA), and PSO. Our simulation results show that the proposed algorithm can cut the computation time down by 51.07% up to 81.77% compared to PSO and GKA.

In this paper, we present an efficient method for speeding up Ant Colony Optimization (ACO), called Pattern Reduction Enhanced Ant Colony Optimization (PREACO). The proposed algorithm is motivated by the observation that many of the computations of ACO on its convergence process are essentially redundant and thus can be eliminated to reduce its computation time. To evaluate the performance of the proposed algorithm, we use it to solve the the traveling salesman problem (TSP). Moreover, we compare the proposed algorithm with several state-of-the-art ACO-based algorithms. Our simulation results indicate that the proposed algorithm can reduce the computation time of ACO algorithms we evaluated up to 99.21% or by a factor of 126.58 while limiting the degradation of the quality of the solution to a very small percentage compared to ACO algorithms themselves.

This paper presents a novel framework based on the notion of pattern reduction, called Framework for Accelerating Metaheuristics via Pattern Reduction (FAMPR), to solve an intrinsic problem of metaheuristics. That is, many computations of metaheuristics during the convergence process are essentially redundant. As such, if they can be eliminated, the computation time of metaheuristics can be significantly reduced. Our experimental result shows that the proposed framework can significantly reduce the computation time of metaheuristics while limiting the loss of quality to a very small percentage.

In this paper, we present a simple but efficient algorithm for integrating a collection of heterogeneous itembanks, called heterogeneous itembanks integrator (HIBI). This algorithm is motivated by the desire to integrate itembanks provided by publishers to the users of an e-learning system, which generally use different content structures. The proposed algorithm starts off with one of the itembanks as a reference itembank. All the items on the other itembanks are integrated into the reference itembank to create the so-called meta-itembank. Moreover, by treating the metaitembank thus created as the reference itembank, it can be easily extended by using exactly the same algorithm. We also use the concept of ontology to share itemsets with other systems. The experimental results showed that the proposed algorithm can provide an extremely high quality result in terms of both the relevance of items and the computation time in Chinese itemsets.

The transmission of multimedia data over WLAN is going to be an unavoidable trend. When packet is transmitted over wireless network, small packet length will increase the header overhead and long packet length will tend to increase the packet error rate. In this article, an adaptive video transmission mechanism under WLAN is proposed. The optimal packet length can be obtained through mathematical analysis according to the factors such as current wireless network transmission status and the header overhead. Based on the optimal packet length, layering and scalability characteristics of MPEG-4 FGS and unequal error protection, a video packet transmission mechanism of low calculation complication is thus proposed. Several distinct scalable video transmission schemes over burst-error channel have been compared, and the simulation results reveal that our transmission scheme is capable of reacting to the varying channel condition with less and smoother quality degradation.

The transmission of WLAN could easily suffer from multi-path or noise interference, which in turn leads to packet loss and the subsequent need of retransmission; moreover, these phenomena could easily lead to the decrease in video transmission quality of some delay-sensitive multimedia applications. Therefore, how to improve the video transmission in WLAN thus becomes a very important issue. In this article, a cross-layer design is proposed to improve video transmission over 802.11e WLANs using error and rate joint control. The network condition estimator (NCE) is designed to distinguish network errors caused by congestion or interference at MAC layer. Based on the information provided from application layer and NCE, the packet-selective rate controller (PSRC) and adaptive TXOP error controller (ATEC) are proposed to eliminate the effect caused by network errors and improve video quality. Simulation results reveal that our cross layer design is capable of reacting to the varying channel condition with less and smoother quality degradation than several distinct cross-layer schemes.

In this paper, we present an Intelligent Cluster Search Engine System, called ICSE. This system is motivated by the observation that traditional search engines present to the users a set of non-classified web pages based on its ranking mechanism, and the unfortunate results are that they usually can not satisfy the need of users. For this reason, ICSE provides to the user a set taxonomic web pages in response to a user's query, and thus it would greatly help the users filter out irrelevant or redundant information. The proposed system can be divided into two parts. The first is the knowledge base constructed by Open Directory Project and Yahoo! Directory. The second is the fast clustering algorithm described herein for clustering the web pages. In addition, in response to a user's query, the proposed system will first send the query to a meta-search engine. Then, it will create a clustered document set using the given knowledge base and the clustering algorithm of ICSE. Our simulation result showed that the proposed system can enhance the relevance and coverage of the search results that the users need compared with traditional search engines.

Most, if not all, of the researches in support vector machine (SVM) based face recognition algorithms have generally presumed that the classifier is static and thus unscalable, due to the fact that SVM is a supervised learning method. This paper introduces a novel SVM based face recognition method - by dynamically adding ¿new¿ faces of existing or new persons into the face database - which circumvents these difficulties. In other words, the proposed algorithm is able to learn and recognize faces that are not in the face database before. The paper presents the theory and the experimental results using the new approach. Our experimental results indicate that the accuracy rate of the proposed algorithm ranges from 91% up to 100% and outperforms all the others.

In this paper, we present a system for predicting the examination result, called automatic score prediction system (ASPS). ASPS builds on the lecture structure created by a teacher or a particular pedagogy a course content tree. Each examination sheet is then derived from this course content tree. Besides, each of the question items in an examination sheet belongs to either a section, a chapter, or a unit The difficulty and authority of the question items are determined by the weighted score. The main task of the system is in the classification of the testing results and in the provision of the learning curves and the remedial courses. The learning achievement and testing results can be summarized as a learning curve to the students. By analyzing the learning accomplishments, not only can ASPS predict the probabilistic result of the next examination, but it can also provide reviews or remedial lessons to the students. The main contributions of the ASPS system described herein are twofold: (1) the system can produce the learning curve based on the examination result, which would make it easy for the students and teachers to understand the learning results. (2) The system can provide the course contents to the students to help them review after the examination. Moreover, it provides suggestions for teaching on the not yet acquainted and confused contents.

Over the past decade, more and more users of the Internet rely on the search engines to help them find the information they need. However, the information they find depends, to a large extent, on the ranking mechanism of the search engines they use. Not surprisingly, it, in general, consists of a large amount of information that is completely irrelevant. To help users of the Internet find the information they are looking for quickly, an efficient algorithm for building the summaries of a collection of documents found by a search engine in response to a user query, called DISCO (Distribution Scoring) is proposed. To demonstrate the performance of the proposed algorithm, Reuters-21578 text categorization collection and some search results from Google are used in our simulation. Moreover, several metrics such as coverage, overlap, and the computation time are employed in evaluating the quality and quantity of the proposed algorithm. All our simulation results indicate that the proposed algorithm outperforms all the existing algorithms in terms of not only the usefulness of the summaries but also the running time.

In this paper, we present a simple but fast codebook generation algorithm, called PREGLA (pattern reduction enhanced GLA). The proposed algorithm is fundamentally different from the previous approaches in that the latter focuses on reducing the size of the codebook whereas the former focuses on using pattern reduction to reduce the computation time. The proposed algorithm is motivated by the observation that input vectors that are ldquostaticrdquo during the training process can be considered as part of the final solutions and thus can be compressed and removed to eliminate the redundant computations at the later iterations of the training process. To evaluate the performance of the proposed algorithm, we compare the proposed algorithm with ldquoefficientrdquo VQ algorithms published recently such as nearest partition set search, fast vector quantization algorithm, and standard GLA. Our experimental results indicate that the proposed algorithm can reduce the computation time from 27.86% up to about 80.45% compared to that of standard GLA and other fast GLA-based algorithms alone.

This paper presents a simple but efficient algorithm for enhancing the performance of GA or GA-based algorithms while retaining the diversity of the search directions. The proposed algorithm is motivated by the observation that some of the genes common to all the individuals during the evolution process can be considered as part of the final solutions and thus can be removed to eliminate the redundant computations at the later generations of the evolution process. To evaluate the performance of the proposed algorithm, we use it to solve the traveling salesman problem (TSP). The benchmarks for the TSP problem range in size from 574 up to 2,152 cities. For the three problems evaluated, our experimental results indicate that the proposed algorithm can reduce the computation time from 28% up to about 84% compared to that of traditional GA and GA-based algorithms alone.

In this paper, we present an efficient algorithm, called pattern reduction (PR) algorithm, to reduce the time required for data clustering based on iterative clustering algorithms. Conceptually similar to a lossy data compression scheme, this algorithm removes at each iteration those data patterns that are close to the centroid of a cluster or remain in the same cluster for a certain number of iterations in a row and are thus unlikely to be moved again from one cluster to another at later iterations by computing a new pattern to represent all the data patterns removed. Our simulation results - from 2 to 1,000 dimensions and 150 to 6,000,000 patterns - indicate that the proposed algorithm can reduce the computation time of k-means, genetic k-means algorithm (GKA) and k-means with genetic algorithm (KGA) from 10% up to about 80% and that for high dimensional data sets, it can even reduce the computation time for more than 70%.

This paper presents a new internet search engine system called document clustering for search engines (DCSE). This system focuses on overcoming the following challenges faced by search engines: (1) relevance of the search results in response to a user query and (2) information coverage. The DCSE system is based upon a meta-search engine that integrates information retrieval (IR), information extraction (IE), genetic algorithm (GA) and document clustering algorithm into a single system. DCSE utilizes information extraction techniques and vector space model (VSM) calculations to determine the relevance of various data, and then categorizes the data via information retrieval and document clustering algorithm in order to better refine the result. Users will receive information that has been calculated and sorted and web links that are ranked according to their relevance. The end result will reduce the amount of time that users spend filtering out irrelevant data.

The paper proposes an intelligent Web portal system (IWPS) as a way to integrate Web page information. Web portals serve as an essential role to the development of the Internet, and offer useful resources and services to large numbers of Web users. Although most Web portals offer complete Web site information, they simply collect and classify Web sites and do not provide a further analysis of the relationship of Web content information. This study aims to develop an intelligent Web portal system that unlike traditional Web portals uses a number of intelligent methods, including information extraction (IE), information retrieval (IR), and heuristic algorithm (HA), and provides more meaningful and integrated information to Web users.

With the development of mobile devices, people can execute many applications on their personal devices. However, many limitations of the mobile devices, e.g., CPU, memory, power supply, etc., make them impossible to be completely the same as the desktop PCs. In this paper we present an integrated management architecture for thin-client purpose called Agent and Profile Management System (APMS). The users only need to download a simple service agent to their mobile device and install it, and then the service agent will connect to the corresponding service provider send a request to the back-end server After receiving the request, the back-end server will execute appropriate processes and return a response to the mobile device. Most of the procedures are accomplished on the server side. Therefore, the workload of mobile devices is relatively low and the cost of mobile devices can be effectively reduced.

In many parallel programs, run-time data redistribution is usually required to enhance data locality and reduce remote memory access on the distributed memory multicomputers. Recently researches in data redistribution algorithm have become very mature. The time required to generate data sets and processor sets is much lesser then before. That means packing/unpacking becomes a relatively heavy cost in the redistribution. In this paper we present methods to perform BLOCK-CYCLIC(s) to BLOCK-CYCLIC(t) redistribution using MPI user-defined types. In this approach, we can reduce the requirement of memory buffers and avoid unnecessary data-movement. The theoretical models are presented to determine the best method for redistribution. To evaluate the performance of the proposed methods, we have implemented our methods on an IBM SP2 parallel machine. The experimental results show that this approach can obviously improve the performance of redistribution in most cases.

Service replication on the server cluster is becoming increasingly important and widespread as the explosive growth of the Web is straining the architecture of many Internet sites. Layer-7 routing (also known as content-aware routing), routing packets based on requested content, has been recognized as a powerful approach to distribute workload among these server farms. However, managing a server cluster with the layer-7 routing mechanism faces numerous challenges. This paper presents a management system that can support a configurable and high-performance layer-7 routing system. In the proposed system, we devise a data structure termed URL table to hold content-related information for making content-aware routing decisions. We also propose a novel idea termed "URL formalization" that provides an efficient solution to speedup the content-aware request routing. The proposed mechanisms are integrated into a management system, which provides a specification interface and enforcement mechanisms so that the system policy can be specified and enforced.

Supporting multicast communications in WDM networks has become an important and attractive topic. We study constrained multicast routing with power consideration in WDM networks, where power is attenuated due to (i) the number of concatenated fanouts (splitting), and (ii) the propagation from source node to any multicast member. We propose a heuristic algorithm called centralized-splitting algorithm to re-construct the multicast routing paths established by the member-only algorithm, which can provide the best bandwidth and wavelength usage in the multicast routing construction with sparse splitting constraint. In comparison to the member-only algorithm, the centralized-splitting algorithm (i) does not increase the time complexity significantly, especially when the size of multicast group is large; (ii) can reduce the power loss up to 17% in average, while it still maintains proper bandwidth, wavelength usage and delay requirements.

Modern Web services must support large and rapidly growing user populations, and remain available 24 hours per day, 7 day per week. Server-cluster is the most promising approach to address this challenge. We further augment the server-cluster schemes with a novel mechanism that enables a Web request to be smoothly migrated and recovered on another working node in the presence of server failure or overload. The new mechanism provides a powerful solution to fault tolerance and dynamic load-distribution for Web services. The administrator can explicitly specify some services to be guaranteed for fault-tolerance or higher performance support. We present the details of our design, implementation, and performance data. The performance results show that the proposed mechanism is efficient and with low associated overhead.

Clusters of commodity computers are becoming an increasingly popular approach for building cost-effective high-performance Internet servers. However, how to place and manage content in such a distributed and complex system is becoming a challenging problem. In particular, such distributed servers tend to be more heterogeneous, and this heterogeneity will further increase the management burden. This paper describes the motivation, design, implementation and performance of a content placement and management system for a heterogeneous distributed Web server.

Clustered server architecture is increasingly being viewed as a successful and cost-effective approach to building a high-performance Web server. Existing server-clustering schemes have typically concentrated on the following issues: scalability, high availability, and user transparency. In this paper, we argue that the design goals of the Web server cluster should include adaptability, fault tolerance, and high manageability. In the presence of the Internet's highly unpredictable workload, the server system should be self-adapting to changing circumstances. We address this problem by building a Web server on a cluster of non-dedicated workstations. Such a server can easily recruit non-dedicated nodes dynamically in response to load bursts. Based on such a scheme, we designed and implemented an innovative approach that enables an ongoing request to be smoothly migrated to another node either in response to a node failure or overload. We also designed and implemented a management system that enables the Web site manager to manage and maintain the distributed server as a single large system.

Clustered server architectures have been employed for many years on the Internet as a way to improve performance, reliability and scalability in the presence of the Internet's explosive growth. Placing and managing content in such a complex system poses a number of challenging problems. In this paper, we describe the work that we are pursuing to design and implement a content placement and management system for the cluster-based Web server.

Today, a successful Internet service is absolutely critical to be up 100 percent of the time. Server clustering is the most promising approach to meet this requirement. However, the existing Web server-clustering solutions merely can provide high availability derived from its redundancy nature, but offer no guarantee about fault resilience for the service. In this paper, we address this problem by implementing an innovative mechanism that enables a Web request to be smoothly migrated and recovered on another working node in the presence of server failure. We will show that the request migration and recovery could be efficiently achieved in the manner of user transparency. The achieved capability of fault resilience is important and essential for a variety of critical services (e.g., E-commerce), which are increasingly widespread used. Our approach takes an important step toward providing a highly reliable Web service.

In many scientific applications, dynamic array redistribution is usually required to enhance the performance of an algorithm. In this paper, we present a generalized basic-cycle calculation (GBCC) method to efficiently perform a BLOCK-CYCLIC(s) over P processors to BLOCK-CYCLIC(t) over Q processors array redistribution. In the GBCC method, a processor first computes the source/destination processor/data sets of array elements in the first generalized basic-cycle of the local array it owns. A generalized basic-cycle is defined as lcm(sP, tQ)/(gcd(s,t)/spl times/P) in the source distribution and lcm(sP, tQ)/(gcd(s,t)/spl times/Q) in the destination distribution. From the source/destination processor/data sets of array elements in the first generalized basic-cycle, we can construct packing/unpacking pattern tables to minimize the data-movement operations. Since each generalized basic-cycle has the same communication pattern, based on the packing/unpacking pattern tables, a processor can pack/unpack array elements efficiently. To evaluate the performance of the GBCC method, we have implemented this method on an IBM SP2 parallel machine, along with the PITFALLS method and the ScaLAPACK method. The cost models for these three methods are also presented. The experimental results show that the GBCC method outperforms the PITFALLS method and the ScaLAPACK method for all test samples. A brief description of the extension of the GBCC method to multidimensional array redistributions is also presented.

In many scientific applications, dynamic array redistribution is usually required to enhance the performance of an algorithm. We present a generalized basic cycle calculation (GBCC) method to efficiently perform a BLOCK-CYCLIC(s) over P processors to BLOCK-CYCLIC(t) over Q processors array redistribution. In the GBCC method, a processor first computes the source/destination processor/data sets of array elements in the first generalized basic cycle of the local array it owns. A generalized basic cycle is defined as lcm(sP,tQ)/(gcd(s,t)/spl times/P) in the source distribution and lcm(sP,tQ)/(gcd(s,t)/spl times/Q) in the destination distribution. From the source/destination processor/data sets of array elements in the first generalized basic cycle, we can construct packing/unpacking pattern tables. Based on the packing/unpacking pattern tables, a processor can pack/unpack array elements efficiently. To evaluate the performance of the GBCC method, we have implemented this method on an IBM SP2 parallel machine, along with the PITFALLS method and the ScaLAPACK method. The cost models for these three methods are also presented. The experimental results show that the GBCC method outperforms the PITFALLS method and the ScaLAPACK method for all test samples. A brief description of the extension of the GBCC method to multi dimensional array redistributions is also presented.