Skip to main content
main-content

Über dieses Buch

This book constitutes the refereed proceedings of the 7th IFIP TC 12 International Conference on Intelligent Information Processing, IIP 2012, held in Guilin, China, in October 2012. The 39 revised papers presented together with 5 short papers were carefully reviewed and selected from more than 70 submissions. They are organized in topical sections on machine learning, data mining, automatic reasoning, semantic web, information retrieval, knowledge representation, social networks, trust software, internet of things, image processing, and pattern recognition.

Inhaltsverzeichnis

Frontmatter

Keynote Presentations

The AI Journey: The Road Traveled and the (Long) Road Ahead

In this talk I will first briefly summarize the many impressive results we have achieved along the road so far traveled in the field of AI including some concrete results obtained at the IIIA-CSIC. Next I will describe some of the future challenges to be faced along the (long) road we still have ahead of us with an emphasis on integrated systems, a necessary step towards human-level AI. Finally I will comment on the importance of interdisciplinary research to build such integrated systems (for instance, sophisticated robots having artificial cartilages, artificial muscles, artificial skin, etc) using some examples related to materials science.

Ramon Lopez de Mantaras

Transfer Learning and Applications

In machine learning and data mining, we often encounter situations where we have an insufficient amount of high-quality data in a target domain, but we may have plenty of auxiliary data in related domains. Transfer learning aims to exploit these additional data to improve the learning performance in the target domain. In this talk, I will give an overview on some recent advances in transfer learning for challenging data mining problems. I will present some theoretical challenges to transfer learning, survey the solutions to them, and discuss several innovative applications of transfer learning, including learning in heterogeneous cross-media domains and in online recommendation, social media and social network mining.

Qiang Yang

Semantics of Cyber-Physical Systems

The very recent development of Cyber-Physical Systems (CPS) provides a smart infrastructure connecting abstract computational artifacts with the physical world. The solution to CPS must transcend the boundary between the cyber world and the physical world by providing integrated models addressing issues from both worlds simultaneously. This needs new theories, conceptual frameworks and engineering practice. In this paper, we set out the key requirements that must be met by CPS systems, review and evaluate the progress that has been made in the development of theory, conceptual frameworks and practical applications. We then discuss the need for semantics and a proposed approach for addressing this. Grand challenges to informatics posed by CPS are raised in the paper.

Tharam Dillon, Elizabeth Chang, Jaipal Singh, Omar Hussain

Big Data Mining in the Cloud

Big Data is the growing challenge that organizations face as they deal with large and fast-growing sources of data or information that also present a complex range of analysis and use problems. Digital data production in many fields of human activity from science to enterprise is characterized by an exponential growth. Big data technologies will become a new generation of technologies and architectures which is beyond the ability of commonly used software tools to capture, manage, and process the data within a tolerable elapsed time.

Massive data sets are hard to understand, and models and patterns hidden within them cannot be identified by humans directly, but must be analyzed by computers using data mining techniques. The world of big data present rich cross-media contents, such as text, image, video, audio, graphics and so on. For cross-media applications and services over the Internet and mobile wireless networks, there are strong demands for cross-media mining because of the significant amount of computation required for serving millions of Internet or mobile users at the same time. On the other hand, with cloud computing booming, new cloud-based cross-media computing paradigm emerged, in which users store and process their cross-media application data in the cloud in a distributed manner. Cross-media is the outstanding characteristics of the age of big data with large scale and complicated processing task. Cloud-based Big Data platforms will make it practical to access massive compute resources for short time periods without having to build their own big data farms. We propose a framework for cross-media semantic understanding which contains discriminative modeling, generative modeling and cognitive modeling. In cognitive modeling, a new model entitled CAM is proposed which is suitable for cross-media semantic understanding. A Cross-Media Intelligent Retrieval System (CMIRS), which is managed by ontology-based knowledge system KMSphere, will be illustrated.

This talk also concerns Cloud systems which can be effectively employed to handle parallel mining since they provide scalable storage and processing services, as well as software platforms for developing and running data analysis environments. We exploit Cloud computing platforms for running big data mining processes designed as a combination of several data analysis steps to be run in parallel on Cloud computing elements. Finally, the directions for further researches on big data mining technology will be pointed out and discussed.

Zhongzhi Shi

Research on Semantic Programming Language

As technologies of Semantic Web Service are gradually matured, developing intelligent web applications with Semantic Web Services becomes an important research topic in Software Engineering. This speech introduces our efforts on Semantic Web Service oriented programming. Employing the concept of semantic computing into service-oriented programming, we proposed a programming language SPL, Semantic Programming Language, which supports the expression and process of semantic information. Based on collaboration of semantic space and information space, the running mechanism of SPL program is presented, which provides SPL program with higher flexibility and stronger adaptability to changes. Furthermore, with the introduction of semantic operators, a kind of searching conditional expression is offered to facilitate the search of Semantic Web Services with greater preciseness and higher flexibility. Besides, semantic based policy and exception mechanism are also brought in to improve the intelligence of policy inference and exception handing in SPL program. At the same time, a platform that supports design and running of SPL program is developed.

Shi Ying

Machine Learning

Effectively Constructing Reliable Data for Cross-Domain Text Classification

Traditional classification algorithms often fail when the

independent

and

identical distributed

(i.i.d.) assumption does not hold, and the cross-domain learning emerges recently is to deal with this problem. Actually, we observe that though the trained model from training data may not perform well over all test data, it can give much better prediction results on a subset of the test data with high prediction confidence. Also this subset of data from test data set may have more similar distribution with the test data. In this study, we propose to construct the reliable data set with high prediction confidence, and use this reliable data as training data. Furthermore, we develop an EM algorithm to refine the model trained from the reliable data. The extensive experiments on text classification verify the effectiveness and efficiency of our methods. It is worth to mention that the model trained from the reliable data achieves a significant performance improvement compared with the one trained from the original training data, and our methods outperform all the baseline algorithms.

Fuzhen Zhuang, Qing He, Zhongzhi Shi

Improving Transfer Learning by Introspective Reasoner

Traditional learning techniques have the assumption that training and test data are drawn from the same data distribution, and thus they are not suitable for dealing with the situation where new unlabeled data are obtained from fast evolving, related but different information sources. This leads to the cross-domain learning problem which targets on adapting the knowledge learned from one or more source domains to target domains. Transfer learning has made a great progress, and a lot of approaches and algorithms are presented. But negative transfer learning will cause trouble in the problem solving, which is difficult to avoid. In this paper we have proposed an introspective reasoner to overcome the negative transfer learning.

Introspective learning exploits explicit representations of its own organization and desired behavior to determine when, what, and how to learn in order to improve its own reasoning. According to the transfer learning process we will present the architecture of introspective reasoner for transductive transfer learning.

Zhongzhi Shi, Bo Zhang, Fuzhen Zhuang

PPLSA: Parallel Probabilistic Latent Semantic Analysis Based on MapReduce

PLSA(Probabilistic Latent Semantic Analysis) is a popular topic modeling technique for exploring document collections. Due to the increasing prevalence of large datasets, there is a need to improve the scalability of computation in PLSA. In this paper, we propose a parallel PLSA algorithm called PPLSA to accommodate large corpus collections in the MapReduce framework. Our solution efficiently distributes computation and is relatively simple to implement.

Ning Li, Fuzhen Zhuang, Qing He, Zhongzhi Shi

Analysis on Limitation Origins of Information Theory

The limitations of Shannon information theory are pointed out from new perspectives. The limitations mainly exist in the neglects of the information reliability and completeness. The significances of the information reliability to the information measurements are further illustrated through example analysis. It is pointed out that such limitations originate from neglects of multi-level information uncertainties, uncertainty of the model and other objects of information system, and insufficient knowledge on uncertainties of probability values.

Yong Wang, Huadeng Wang, Qiong Cao

Data Mining

Intelligent Inventory Control: Is Bootstrapping Worth Implementing?

The common belief is that using Reinforcement Learning methods (RL) with bootstrapping gives better results than without. However, inclusion of bootstrapping increases the complexity of the RL implementation and requires significant effort. This study investigates whether inclusion of bootstrapping is worth the effort when applying RL to inventory problems. Specifically, we investigate bootstrapping of the temporal difference learning method by using eligibility trace. In addition, we develop a new bootstrapping extension to the Residual Gradient method to supplement our investigation. The results show questionable benefit of bootstrapping when applied to inventory problems. Significance tests could not confirm that bootstrapping had statistically significantly reduced costs of inventory controlled by a RL agent. Our empirical results are based on a variety of problem settings, including demand correlations, demand variances, and cost structures.

Tatpong Katanyukul, Edwin K. P. Chong, William S. Duff

Support Vector Machine with Mixture of Kernels for Image Classification

Image classification is a challenging problem in computer vision. Its performance heavily depends on image features extracted and classifiers to be constructed. In this paper, we present a new support vector machine with mixture of kernels (SVM-MK) for image classification. On the one hand, the combined global and local block-based image features are extracted in order to reflect the intrinsic content of images as complete as possible. SVM-MK, on the other hand, is constructed to shoot for better classification performance. Experimental results on the Berg dataset show that the proposed image feature representation method together with the constructed image classifier, SVMMK, can achieve higher classification accuracy than conventional SVM with any single kernels as well as compare favorably with several state-of-the-art approaches.

Dongping Tian, Xiaofei Zhao, Zhongzhi Shi

The BDIP Software Architecture and Running Mechanism for Self-Organizing MAS

As there are huge gaps between the local micro interactions among agents and the global macro emergence of self-organizing system, it is a great challenge to design self-organizing mechanism and develop self-organizing multi-agent system to obtain expected emergence. Policy-based self-organization approach is helpful to deal with the issue, in which policy is the abstraction of self-organizing mechanism and acts as the bridge between the local micro interactions and global macro emergence. This paper focuses on how to develop software agents in policy-based self-organizing multi-agent system and proposes a BDIP architecture of software agent. In our approach, policy is an internal component that encapsulates the self-organizing information and integrates with BDI components. BDIP agent decides its behaviors by complying with the policies and respecting BDI specifications. An implementation model and the running mechanism as well as corresponding decision algorithms for BDIP agents are studied. A case of self-organizing system is studied to illustrate our proposed approach and show its effectiveness.

Yi Guo, Xinjun Mao, Fu Hou, Cuiyun Hu, Jianming Zhao

Optimization of Initial Centroids for K-Means Algorithm Based on Small World Network

K-means algorithm is a relatively simple and fast gather clustering algorithm. However, the initial clustering center of the traditional k-means algorithm was generated randomly from the dataset, and the clustering result was unstable. In this paper, we propose a novel method to optimize the selection of initial centroids for k-means algorithm based on the small world network. This paper firstly models a text document set as a network which has small world phenomenon and then use small-world’s characteristics to form k initial centroids. Experimental evaluation on documents croups show clustering results (total cohesion, purity, recall) obtained by proposed method comparable with traditional k-means algorithm. The experiments show that results are obtained by the proposed algorithm can be relatively stability and efficiency. Therefore, this method can be considered as an effective application in the domain of text documents, especially in using text clustering for topic detection.

Shimo Shen, Zuqiang Meng

ECCO: A New Evolutionary Classifier with Cost Optimisation

Decision tree learning algorithms and their application represent one the major successes of AI. Early research on these algorithms aimed to produce classification trees that were accurate. More recently, there has been recognition that in many applications, aiming to maximize accuracy alone is not adequate since the cost of misclassification may not be symmetric and that obtaining the data for classification may have an associated cost. This has led to significant research on the development of cost-sensitive decision tree induction algorithms. One of the seminal studies in this field has been the use of genetic algorithms to develop an algorithm known as ICET. Empirical trials have shown that ICET produces some of the best results for cost-sensitive decision tree induction. A key feature of ICET is that it uses a pool that consists of genes that represent biases and parameters. These biases and parameters are then passed to a decision tree learner known as EG2 to generate the trees. That is, it does not use a direct encoding of trees. This paper develops a new algorithm called ECCO (Evolutionary Classifier with Cost Optimization) that is based on the hypothesis that a direct representation of trees in a genetic pool leads to improvements over ICET. The paper includes an empirical evaluation of this hypothesis on four data sets and the results show that, in general, ECCO is more cost-sensitive and effective than ICET when test costs and misclassifications costs are considered.

Adam Omielan, Sunil Vadera

Automatic Reasoning

Reasoning Theory for D3L with Compositional Bridge Rules

The semantic mapping in Distributed Dynamic Description Logics (D3L) allows knowledge to propagate from one ontology to another. The current research for knowledge propagation in D3L is only for a simplified case when only two ontologies are involved. In this paper we study knowledge propagation in more complex cases. We find in the case when more than two ontologies are involved and bridge rules form chains, knowledge does not always propagate along chains of bridge rules even if we would expect it. Inspired by Package-based description Logics, we extend the original semantics of D3L by imposing so called compositional consistency condition on domain relations in D3L interpretations. Under this semantics knowledge propagates along chains of bridge rules correctly. Furthermore we provide a distributed Tableaux reasoning algorithm for deciding satisfiability of concepts which is decidable in D3L under compositional consistency. Compared with original one, the extended D3L provides more reasonable logic foundation for distributed, dynamic system such as the information integration system and the Semantic Web.

Xiaofei Zhao, Dongping Tian, Limin Chen, Zhongzhi Shi

Semantic Keyword Expansion: A Logical Approach

Keyword search is the primary way for ordinary users to access the Web content. It is essentially a syntax match between users’ key-ins and the index structure of information systems with a relevance-ranking way to sort the hitted documents. The syntax match can rarely satisfy users’ information need when the keywords and index are not syntically similar while share a lot semantically. This paper proposed a way of semantic keyword expansion with a re-ranking to handle this problem. Experimental results show that our method helps in improving the quality of keyword search and particularly in the cases of keywords with widely-used synonyms or parasynonyms.

Limin Chen

An ABox Abduction Algorithm for the Description Logic ALCI

ABox abduction is the foundation of abductive reasoning in description logics. It finds the minimal sets of ABox axioms which could be added to a background knowledge base to enforce the entailment of certain ABox assertions. In this paper, an abductive reasoning algorithm for the description logic ALCI is presented. The algorithm is an extension of an existing ABox abduction algorithm for the description logic ALC, with the feature that it is based on the Tableau of ALCI directly and do not need to use arguments and Skolem terms. It firstly transforms the ABox abduction problem into the consistency problem of knowledge base; then traditional Tableau construction process for ALCI is expanded to deal with this problem; finally the solution of the abduction problem is constructed by a process of backtracking.

Yanwei Ma, Tianlong Gu, Binbin Xu, Liang Chang

Reasoning about Assembly Sequences Based on Description Logic and Rule

Reasoning about assembly sequences is useful for identifying the feasibility of assembly sequences according to the assembly knowledge. Technologies used for reasoning about assembly sequences have crucial impacts on the efficiency and automation of assembly sequence planning. Description Logic (DL) is well-known for representing and reasoning about knowledge of static application domains; it offers considerable expressive power going far beyond propositional logic while reasoning is still decidable. In this paper, we bring the power and character of description logic into reasoning about assembly sequences. Assembly knowledge is firstly described by a description logic enhanced with some rules. Then, the feasibility of assembly operations is decided by utilizing the reasoning services provided by description logics and rules. An example has been provided to demonstrate the usefulness and executability of the proposed approach.

Yu Meng, Tianlong Gu, Liang Chang

Semantic Web

Dynamic Logic for the Semantic Web

The propositional dynamic logic PDL is one of the most successful variants of modal logic; it plays an important role in many fields of computer science and artificial intelligence. As a logical basis for the W3C-recommended Web ontology language OWL, description logic provides considerable expressive power going far beyond propositional logic as while as the reasoning is still decidable. In this paper, we bring the power and character of description logic into PDL and present a dynamic logic

ALC

-DL for the semantic Web. The logic

ALC

-DL inherits the knowledge representation ability of both the description logic ALC and the logic PDL. With an approach based on Buchi tree automaton, we prove that the satisfiability problem of

ALC

-DL formulas is still decidable and is EXPTIME-complete. The logic

ALC

-DL is suitable for modeling and reasoning about dynamic knowledge in the semantic Web environment.

Liang Chang, Qicheng Zhang, Tianlong Gu, Zhongzhi Shi

On the Support of Ad-Hoc Semantic Web Data Sharing

Sharing Semantic Web datasets provided by different publishers in a decentralized environment calls for efficient support from distributed computing technologies. Moreover, we argue that the highly dynamic ad-hoc settings that would be pervasive for Semantic Web data sharing among personal users in the future pose even more demanding challenges for enabling technologies. We propose an architecture that is based upon the peer-to-peer (P2P) paradigm for ad-hoc Semantic Web data sharing and identify the key technologies that underpin the implementation of the architecture. We anticipate that our (current and future) work will offer powerful support for sharing of Semantic Web data in a decentralized manner and becomes an indispensable and complementary approach to making the Semantic Web a reality.

Jing Zhou, Kun Yang, Lei Shi, Zhongzhi Shi

An Architecture Description Language Based on Dynamic Description Logics

ADML is an architectural description language based on Dynamic Description Logic for defining and simulating the behavior of system architecture. ADML is being developed as a new formal language and/or conceptual model for representing the architectures of concurrent and distributed systems, both hardware and software. ADML embraces dynamic change as a fundamental consideration, supports a broad class of adaptive changes at the architectural level, and offers a uniform way to represent and reason about both static and dynamic aspects of systems. Because the ADML is based on the Dynamic Description Logic DDL(

$\mathcal{SHON}$

(D)), which can represent both dynamic semantics and static semantics under a unified logical framework, architectural ontology entailment for the ADML languages can be reduced to knowledge base satisfiability in DDL(

$\mathcal{SHON}$

(D)), and dynamic description logic algorithms and implementations can be used to provide reasoning services for ADML. In this article, we present the syntax of ADML, explain its underlying semantics using the Dynamic Description Logic DDL(

$\mathcal{SHON}$

(D)), and describe the core architecture description features of ADML.

Zhuxiao Wang, Hui Peng, Jing Guo, Ying Zhang, Kehe Wu, Huan Xu, Xiaofeng Wang

Information Retrieval

Query Expansion Based-on Similarity of Terms for Improving Arabic Information Retrieval

This research suggests a method for query expansion on Arabic Information Retrieval using Expectation Maximization (EM). We employ the EM algorithm in the process of selecting relevant terms for expanding the query and weeding out the non-related terms. We tested our algorithm on INFILE test collection of CLLEF2009, and the experiments show that query expansion that considers similarity of terms both improves precision and retrieves more relevant documents. The main finding of this research is that we can increase the recall while keeping the precision at the same level by this method.

Khaled Shaalan, Sinan Al-Sheikh, Farhad Oroumchian

Towards an Author Intention Based Computational Model of Story Generation

This paper addresses the problem of plot controllable story generation model. Since most of the previous works focus on logically flawless plot, the main challenge is the need for more generic method to generate dramatic and interesting plot. Motivated by this background, this paper proposes a computational method for plot controllable story generation. Firstly, we use planning as a model of plot generation. Then we utilize author intentions as plot constraints to force the planner to consider substantially more complex plans. Finally we integrate author intentions into planning and develop a plot controllable Graphplan algorithm. Experimental results demonstrate the effectiveness of our approach.

Feng Zhu, Cungen Cao

Adaptive Algorithm for Interactive Question-Based Search

Popular web search engines tend to improve the relevance of their result pages, but the search is still keyword–oriented and far from ”understanding” the queries’ meaning. In the article we propose an interactive question-based search algorithm that might come up helpful for identifying users’ intents. We describe the algorithm implemented in a form of a questions game. The stress is put mainly on the most critical aspect of this algorithm – the selection of questions. The solution is compared to broadly used decision trees learning algorithms in terms of knowledge inconsistencies tolerance, efficiency and complexity.

Jacek Rzeniewicz, Julian Szymański, Włodzisław Duch

Research of Media Material Retrieval Scheme Based on XPath

With more and more media materials appear on the internet, it comes a sharp problem of how to manage these resources and how to search them efficiently. We construct a management system of media material for the reliable wideband network. According to a detailed analysis of the characteristic of media material queries, we proposed a hierarchical indexing mechanism based on XPath language to discover the resources that match a given query. Our system permits users to locate data iteratively even using scarce information. The description of materials is mapped onto the DHT index. Our indexing scheme has good properties such as space efficient, good scalability and resilient to arbitrary linking.

Shuang Feng, Weina Zhang

Construction of SCI Publications Information System for Statistic

There are over 8000 SCI (Science Citation Index) publications in the ISI (Institute for Scientific Information) Web of Knowledge database system. However, the publications are too many and it is difficult for new authors to choose the most suitable journals or periodicals to submit their research fruits of high level. So, some valuable information about SCI publications is collected, and the corresponding database is established. The records from this database are classified and counted. The statistical results show that the SCI publications information system is helpful to authors to issue papers.

Xie Wu, Huimin Zhang, Jingbo Jiang

Knowledge Representation

Symbolic ZBDD Representations for Mechanical Assembly Sequences

The representations of assembly knowledge and assembly sequences are crucial in assembly planning, where the size of parts involved is a significant and often prohibitive difficulty. Zero-suppressed binary decision diagram (ZBDD) is an efficient form to represent and manipulate the sets of combination, and appears to give improved results for large-scale combinatorial optimization problems. In this paper, liaison graphs, translation functions, assembly states and assembly tasks are represented as sets of combinations, and the symbolic ZBDD representation of assembly sequences is proposed. An example is given to show the feasibility of the ZBDD-based representation scheme.

Fengying Li, Tianlong Gu, Guoyong Cai, Liang Chang

The Representation of Indiscernibility Relation Using ZBDDs

The indiscernibility relation is the basic concept in Rough set theory, a novel representation of indiscernibility relation using Zero-Suppressed BDDs is proposed in this paper. Through introducing the indiscernibility matrix and the indiscernibility graph, we put forward the encoding of the variable and give the characteristic function. Once the characteristic function is constructed, it can be represented using ZBDDs.And further, combined with an example, we analyze the effectiveness of this method. It provides a basis for deal with rough set computing.

Qianjin Wei, Tianlong Gu, Fengying Li, Guoyong Cai

Symbolic OBDD Assembly Sequence Planning Algorithm Based on Unordered Partition with 2 Parts of a Positive Integer

To improve solution efficiency and automation of assembly sequence planning, a symbolic ordered binary decision diagram (OBDD) technique for assembly sequence planning problem based on unordered partition with 2 parts of a positive integer is proposed. To convert the decomposition of assembly liaison graph into solving unordered partition with 2 parts of positive integer

N

, a transformation method from subassembly of the assembly to positive integer

N

is proposed, the judgment methods for the connectivity of a graph and geometrical feasibility of each decomposition combined with symbolic OBDD technique is proposed too, and all geometrically feasible assembly sequences are represented as OBDD-based AND/OR graph. Some applicable experiments show that the symbolic OBDD based algorithm can generate feasible assembly sequences correctly and completely.

Zhoubo Xu, Tianlong Gu, Rongsheng Dong

A Representation Model of Geometrical Tolerances Based on First Order Logic

Tolerance representation models are used to specify tolerance types and explain semantics of tolerances for nominal geometry parts. To well explain semantics of geometrical tolerances, a representation model of geometrical tolerances based on First Order Logic (FOL) is presented in this paper. We first investigate the classifications of feature variations and give the FOL representations of them based on these classifications. Next, based on the above representations, we present a FOL representation model of geometrical tolerances. Furthermore, we demonstrate the effectiveness of the representation model by specifying geometrical tolerance types in an example.

Yuchu Qin, Yanru Zhong, Liang Chang, Meifa Huang

Social Networks

Modeling Group Emotion Based on Emotional Contagion

Using computer to generate crowd animation to understand the behavior choice and making decision of individuals in crowd has become a trend in several fields. And group emotions have a great impact on group behaviors and group outcomes. Based on the researches of group emotions described by Hatifield, we propose a quantitative method to building the group emotion model which is focus on a group or a crowd, not on an individual. Our aim is to reflect more believable emotion experiences of individuals in social situation; the individuals’ emotion is coming not only from the external stimulus, but from others group members through the conscious and unconscious induction of emotion states as well. For the emotional contagion plays a significant role in the development of group emotion, we select personality, emotional expressivity and susceptibility as mainly factors which influence the intensity of group emotions. Simulation is done by using Netlogo software, and the results show that the model is available and embody the fundamental characteristics of group emotion, and virtual individuals in crowd can generate credible emotional experience and response.

Yanjun Yin, Weiqing Tang, Weiqing Li

Hierarchical Overlapping Community Discovery Algorithm Based on Node Purity

A hierarchical overlapping community discovery algorithm based on node purity (OCFN-PN) is proposed in the paper. This algorithm chooses the maximal relative centrality as the initial community, which solves the problem of inconsistent results of the community discovery algorithm based on fitness resulting from randomly choosing nodes. Before optimizing and merging communities, the community overlapping degree and the joint-union should be calculated so that the problems of twice merging can be solved. Research results show that this algorithm has lower time complexity and the communities obtained by this algorithm are more suitable for real world networks.

Guoyong Cai, Ruili Wang, Guobin Liu

Finding Topic-Related Tweets Using Conversational Thread

Microblog has gained more and more users around the world, the popularity of which makes information spreading in microblog the most important and influential activities on the Internet. Therefore, search in microblog is of the most significant issue for both academic and industrial world. Search in webpages has been studied for several decades, but as for microblog it is still an open and brand new question for everyone. Search in microblog is more difficult than that in traditional webpages because of the sparseness of the messages. Search functions in current microblogging services simply match microblog messages with query words, which cannot guarantee the correlation between the retrieving messages and the users’ intention. We introduce the concept of conversational thread to gain more information and improve the search result in microblog. We also use SVMRank to train a model to determine the rank of relevance of the queries and messages. Through a series of experiments, we proved that our method is easy to implement, and can improve the precision up to 29% in average.

Peng Cao, Shenghua Liu, Jinhua Gao, Huawei Shen, Jingyuan Li, Yue Liu, Xueqi Cheng

Messages Ranking in Social Network

Nowadays, people engage more and more in social networks, such as Twitter, FaceBook, and Orkut, etc. In these social network sites, users create relationship with familiar or unfamiliar persons and share short messages between each other. One arising problem for these social networks is information is real-time and updates so quickly that users often feel lost in such huge information flow and struggle to find what really interest them. In this paper, we study the problem of personalized ranking in social network and use the SSRankBoost algorithm, a kind of pairwise learning to rank method to solve this problem. We evaluate our approach using a real microblog dataset for experiment and analyze the result empirically. The result shows clear improvement compared to those without ranking criteria.

abstract

environment.

Bo Li, Fengxian Shi, Enhong Chen

Trust Software

Diagnosis of Internetware Systems Using Dynamic Description Logic

This paper proposes a kind of multi-agent based internetware system architecture, and introduces a diagnoser that can perform on-line diagnosis by observing the behaviors of it. This diagnoser brings semantic description to the procedure of states conversation of the system to be diagnosed by using dynamic description logic, which makes it possible to make use of more knowledge to analyze the failure further.

Kun Yang, Weiqun Cui, Junheng Teng, Chenzhe Hang

Reasoning about Semantic Web Services with an Approach Based on Temporal Description Logic

Temporal description logic

ALC

-LTL not only has considerable expressive power, but also extends the description capability of description logic from the static domain to the dynamic domain. In this paper,

ALC

-LTL is applied for the composition of semantic Web services. We take the view that atomic process and composite process in the OWL-S ontology can be considered as atomic service and composited service respectively. Inputs, outputs, local variables, preconditions and results of atomic processes can all be described with

ALC

-LTL. Based on the models of services, the executability problem and the projection problem of Web services can be reasoned about effectively.

Juan Wang, Liang Chang, Chuangying Zhu, Rongsheng Dong

Constraint Programming-Based Virtual Machines Placement Algorithm in Datacenter

As underlying infrastructure of cloud computing platform, datacenter is seriously underutilized, however, its operating costs is high. In this paper, we implement virtual machines placement algorithm in CloudSim using constraint programming approach. We first formulate the problem of virtual machines placement in virtualized datacenters as a variant of multi-dimensions bin packing problem, and then exploit constraint solver to solve this problem with the objective of minimizing number of physical machines that host virtual machines. Finally, we compare different virtual placement algorithms for evaluating constraint programming-based virtual machine placement algorithm including the built-in virtual machine placement algorithm in CloudSim and FFD algorithm. The experimental results show that constraint programming-based virtual machines placement algorithm can efficiently reduce the number of physical machines to achieve the goal of reducing datacenter operating costs and improving resource utilization.

Yonghong Yu, Yang Gao

Recommendation-Based Trust Model in P2P Network Environment

Two kinds of peers’ relationship are usually considered for reputation management in P2P network. One of them is direct trust relationship that the reputation is got with two peers interacting directly; the other is recommendation trust that the reputation is got by recommendation of the third party. In this paper, we presented an improved trust model based on recommended in P2P environment. In order to weight the transaction reputation and recommendation reputation, we introduced a risk value which resists the influence of false recommendation and collaborative cheating from malicious peers. The global reputation value of target peer can be calculated by using the different portions of transaction reputation, recommendation reputation and risk value. At last, we made the simulation experiments verifying the ability of resisting threats such as slandered by malicious peers, collaborative cheating, and so on.

Yueju Lei, Guangxi Chen

Internet of Things

Frequency-Adaptive Cluster Head Election in Wireless Sensor Network

Efficient information routing mechanism is a critical research issue for wireless sensor networks (

WSN

) due to the limit energy and storage resource of sensor nodes. The clustering-based approaches (e.g.

LEACH

,

Gupta

and

CHEF

) for information routing have been developed. Although these approaches did actually improve the routing efficiency and prolong the network lifetime, the clustering frequency

f

is usually pre-designed and fixed. However, the network context (e.g. Energy and load) often dynamically changes, which can provide the dynamical adjustment determination for

f

. Hence, in this paper, we propose a frequency-adaptive cluster-head election approach, which applies the network context for making corresponding

f

adjustment. Furthermore, an

f

-based clustering algorithm is presented as well. The case study and experimental evaluations demonstrate the effectiveness of the proposed approach.

Tianlong Yun, Wenjia Niu, Xinghua Yang, Hui Tang, Song Ci

A Cluster-Based Multilevel Security Model for Wireless Sensor Networks

Wireless sensor network is one of the fundamental components of the Internet of Things. With the growing use of wireless sensor networks in commercial and military, data security is a critical problem in these applications. Considerable security works have been studied. However, the majority of these works based on the scenarios that the sensitivities of data in the networks are in the same. In this paper, we present a cluster-based multilevel security model that enforces information flow from low security level to high security level. The design of the model is motivated by the observation that sensor nodes in numerous applications have different security clearances. In these scenarios, it is not enough for just protecting the data at a single level. The multilevel security mechanism is needed to prevent the information flow from high level nodes to low level nodes. We give the formal description of the model and present a scheme to achieve it. In our model, sensor nodes are grouped into different clusters. In each cluster, the security clearance of sensor nodes must not be higher than the security clearance of the cluster head. We use cryptography techniques to enforce the information flow policy of this model. The higher level nodes can derive the keys of lower level nodes and use the derived key to get the information from lower-level nodes.

abstract

environment.

Chao Lee, Lihua Yin, Yunchuan Guo

A New Security Routing Algorithm Based on MST for Wireless Sensor Network

In order to solve the general problems of information overlap, low energy utilization rate and network transmission security in routing protocols of wireless sensor networks, a scheme of energy-efficient and security-high routing for wireless sensor networks (EEASHR) based on improved Kruskal algorithms is proposed in this paper. The proposed scheme takes the energy size required for transmission and the value of reliability between nodes and nodes as the edges value of graph, then uses improved Kruskal algorithm to generate the minimum spanning tree (MST) in sink node, in other words, the optimal route path. The simulation in NS2 shows that the proposed algorithm improves network energy efficiency, reduces the node packet loss rate and prolongs the life cycle of wireless sensor networks.

Meimei Zeng, Hua Jiang

Image Processing

A Novel Model for Semantic Learning and Retrieval of Images

In this paper, we firstly propose an extended probabilistic latent semantic analysis (PLSA) to model continuous quantity. In addition, corresponding EM algorithm is derived to determine the parameters. Then, we apply this model in automatic image annotation. In order to deal with the data of different modalities according to their characteristics, we present a semantic annotation model which employs continuous PLSA and traditional PLSA to model visual features and textual words respectively. These two models are linked with the same distribution over all aspects. Furthermore, an asymmetric learning approach is adopted to estimate the model parameters. This model can predict semantic annotation well for an unseen image because it associates visual and textual modalities more precisely and effectively. We evaluate our approach on the Corel5k and Corel30k dataset. The experiment results show that our approach outperforms several state-of-the-art approaches.

Zhixin Li, ZhiPing Shi, ZhengJun Tang, Weizhong Zhao

Automatic Image Annotation and Retrieval Using Hybrid Approach

We firstly propose continuous probabilistic latent semantic analysis (PLSA) to model continuous quantity. In addition, corresponding Expectation-Maximization (EM) algorithm is derived to determine the model parameters. Furthermore, we present a hybrid framework which employs continuous PLSA to model visual features of images in generative learning stage and uses ensembles of classifier chains to classify the multi-label data in discriminative learning stage. Since the framework combines the advantages of generative and discriminative learning, it can predict semantic annotation precisely for unseen images. Finally, we conduct a series of experiments on a standard Corel dataset. The experiment results show that our approach outperforms many state-of-the-art approaches.

Zhixin Li, Weizhong Zhao, Zhiqing Li, Zhiping Shi

Double Least Squares Pursuit for Sparse Decomposition

Sparse decomposition has been widely used in numerous applications, such as image processing, pattern recognition, remote sensing and computational biology. Despite plenty of theoretical developments have been proposed, developing, implementing and analyzing novel fast sparse approximation algorithm is still an open problem. In this paper, a new pursuit algorithm Double Least Squares Pursuit (DLSP) is proposed for sparse decomposition. In this algorithm, the support of the solution is obtained by sorting the coefficients which are calculated by the first Least-Squares, and then the non-zero values over this support are detected by the second Least-Squares. The results of numerical experiment demonstrate the effectiveness of the proposed method, which is with less time complexity, more simple form, and gives close or even better performance compared to the classical Orthogonal Matching Pursuit (OMP) method.

Wanyi Li, Peng Wang, Hong Qiao

Ensemble of k-Labelset Classifiers for Multi-label Image Classification

In the real world, images always have several visual objects instead of only one, which makes it difficult for traditional object recognition methods to deal with them. In this paper, we propose an ensemble method for multi-label image classification. First, we construct an ensemble of

k

-labelset classifiers. A voting technique is then employed to make predictions for images based on the created ensemble of

k

-labelset classifiers. We evaluate our method on Corel dataset and demonstrate the precision, recall and

F

1

measure superior to the state-of-the-art methods.

Dapeng Zhang, Xi Liu

Pattern Recognition

Robust Palmprint Recognition Based on Directional Representations

In this paper, we consider the common problem of automatically recognizing palmprint with varying illumination and image noise. Gabor wavelets can be well represented for biometric image for their similar characteristics to human visual system. However, these Gabor-based algorithms are not robust for image recognition under non-uniform illumination and noise corruption. To improve the recognition performance under the low quality conditions, we propose novel palmprint recognition approach using directional representations. Firstly, the directional representation for palmprint appearance is obtained by the anisotropy filter, which is robust to drastic illumination changes and preserves important discriminative information. Then, the PCA is employed to reduce the dimension of image feature. At last, based on a sparse representation on palmprint feature, the compressed sensing is used to distinguish palms from different hands. Experimental results on the PolyU palprint database show the proposed algorithm have better performance. And the proposed scheme is robust to varying illumination and noise corruption.

Hengjian Li, Lianhai Wang, Zutao Zhang

FPGA-Based Image Acquisition System Designed for Wireless

Introduced micro-wireless transmission system image acquisition the overall structure, each part of the overall design ideas and works theory. Designed SBBC camera interface and data cache control module, analysis of the TCP / IP communication protocol standard structure and characteristics, and on this basis, the design of a non-standard short-range wireless communication protocol. System which controlled by the input and output module is based on FPGA, the system collected data alternately to the outer SDRAM, this realized the image data collection; Ethernet interface controller also designed in FPGA to achieve the image data transmit through wireless. The results show that: under normal conditions, the wireless communication module can be complete, accurate, stable wireless data transceiver functions.

Haohao Yuan, Jianhe Zhou, Suqiao Li

A Context-Aware Multi-Agent Systems Architecture for Adaptation of Autonomic Systems

An important requirement of autonomic systems is that they self-adapt, both with respect to internal self-healing and with respect to external environmental changes. In order to fulfill this requirement autonomic systems must have awareness abilities, context gathering mechanisms, context-dependent adaptation policies, and the ability to react with respect to each adaptation requirement. In this paper we provide an insight into these issues and propose a multi-agent system that interact faithfully among themselves in order to self adapt with safety and security.

Kaiyu Wan, Vasu Alagar

Eyes Closeness Detection Using Appearance Based Methods

Human eye closeness detection has gained wide applications in human computer interface designation, facial expression recognition, driver fatigue detection, and so on. In this work, we present an extensive comparison on several state of art appearance-based eye closeness detection methods, with emphasize on the role played by each crucial component, including geometric normalization, feature extraction, and classification. Three conclusions are highlighted through our experimental results: 1) fusing multiple cues significantly improves the performance of the detection system; 2) the AdaBoost classifier with difference of intensity of pixels is a good candidate scheme in practice due to its high efficiency and good performance; 3) eye alignment is important and influences the detection accuracy greatly. These provide useful lessons for the future investigations on this interesting topic.

Xue Liu, Xiaoyang Tan, Songcan Chen

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise