Abstract:Markov logic networks (MLNs) is a statistical relational learning (SRL) model, which combines Markov network and first order logic. It has been applied widely in nature language processing, complex networks, information extraction, etc. This paper addresses the theoretical model of Markov logic networks (MLNs), weight and structure learning of MLNs comprehensively, and finally presents future works of MLNs.

Abstract:Anchor texts have been extensively used and have proven to be effective over the years in commercial Web search engines; however, anchor texts are created freely without any supervision, which causes there to be a large number of noises and spam in anchor texts. Moreover, for transactional queries, which need measurements of service quality, destination, Web pages of anchor texts are not always consistent with user experience. To overcome these problems, this paper focuses on the large scale of the user Web browsing behavior data. This paper first proposes a framework for a retrieval effectiveness measurement. Then, based on this framework, analyze the relation between the user Web browsing behavior data and retrieval effectiveness of anchor texts is analyzed. The properties of a user Web browsing behavior, which are useful for Web search, are mined and quantified. Experimental results show that the proposed method offers an accurate evaluation on the retrieval effectiveness of anchor texts.

Abstract:This paper explores semantic role labeling (SRL) in the Chinese language for nominal predicates. In addition to the widely adopted features of verbal SRL, various nominal predicate-specific features are also explored. Moreover, the nominal SRL performance has been improved by properly integrating features that were derived from a state-of-the-art verbal SRL system. Finally, the paper explains in detail the nominal predicate recognition, which is essential in a fully automatic nominal SRL system. Evaluations on Chinese NomBank show that proper integration of a verbal SRL system significantly improves the performance of a nominal SRL. It also shows that this nominal SRL system achieves the performance of 72.67 in F1-measure on golden parse trees and golden predicates, and outperforms the state-of-the-art nominal SRL systems in the Chinese language; however, the performance drops to 55.14 in F1-measure on automatic parse trees and automatic predicates.

Abstract:A dynamic niche-based self-organizing learning algorithm (DNSLA) was proposed in this paper. The dynamic learning mechanism based on 0-1 coding method was carried out, and the individuals involved in this algorithm were able to adapt to the dynamic environments through active learning, which was different from the passive adaptive search strategy in traditional evolutionary algorithms. As a result of self-organizing learning and friendly interaction with the environments, DNSLA was more robust to adapt to the dynamic problems, and it was able to accurately detect the slight changes of the environments and track the extreme points in the solution domain. A series of dynamic simulation tests for comparative experiments showed that, even in the turbulent environments, DNSLA was still able to perform friendly interactive learning with the dynamic environments. DNSLA showed a strong robustness in the comparative experiments, whose dynamic search capabilities were far superior to other search methods.

Abstract:To improve the accuracy of mining results, this paper proposes a method of privacy preserving frequent pattern mining, based on sample learning and synchronized randomization of itemset (LS-PPFM). The method utilizes the data of individuals who do not care privacy. First, the data that does not need protecting are learned, and some strongly associated items are obtained. Then, when the data is randomized, the associated items are bound together and randomized synchronously to try to keep their potential associations. Experimental results show that compared with independent randomization, LS-PPFM can achieve significant improvements on accuracy, while losing a little privacy.

Abstract:This paper characterizes the local and global hotspots in text streams and elaborates their correlation. The paper then applies Kolmogorov complexity to mining the hotspots in multiple text streams. The Redundant Information is defined based on Kolmogorov complexity, and it has been demonstrated that the Redundant Information exceeding a threshold is necessary for the local hotspots. Secondly, a similarity metric, termed as Stream Information Distance (SID), is suggested based on the conditional Kolmogorov complexity to quantify the similarity between different text streams. Borrowing ideas of Phylogeny originated from Computational Biology, a heuristic algorithm based on hierarchical clustering is proposed to mine the global hostspots from multiple text streams. Finally, the convergency, effectiveness, and scalability of this algorithm are validated by the extensive experiments over synthetic and real data set.

Abstract:Search engine queries are often too vague to achieve relevant results. This paper presents an intelligent query approach that can distinguish vague queries and organize the related queries of each vague query into a concept hierarchy. Through the concept hierarchy, users can quickly find proper queries for their informational needs. The TECH (term concept hunting) is proposed, based on the small world of human languages. TECH utilizes both the community detection algorithms in the physical field and IR techniques in the computer science field to generate an extensible framework. Experimental results show that compared with the traditional listing query suggestion manner, users prefer the intelligent query suggestion. TECH can effectively distinguish vague queries and significantly outperforms the other three state-of-the-art hierarchical building systems statistically.

Abstract:This paper presents a general detection framework, and develops a variety of content and structure features to find high quality threads. The feature selection algorithm, which is a combination of genetic algorithm, Tabu search and a machine learning algorithm, is designed to attain a better assessment of key features. In this paper, an experiment is done that focuses on the Tencent Message Boards. The experimental results, obtained from a large scale evaluation of over thousands of real web forum threads and user ratings, demonstrate the feasibility of modeling and detecting high quality threads. The proposed feature extraction methods, feature selection algorithms, and detection framework can be useful for a variety of domains such as Blogs and social network platforms.

Abstract:To solve the problem of multiple continuous K nearest neighbor (KNN) queries over moving objects, considering the development of multi-core and multi-threading technologies, a two-stage framework is proposed for Multi-Threading Processing of Multiple Continuous KNN Queries (MPMCQ). This includes a preprocessing stage and a query execution stage to carry out the data updating task and the query execution task separately. In each of the stages, techniques are designed to optimize the cache access hit ratio and improve the parallelism through multi-threading. A query grouping technique in the query execution stage is proposed to improve the data temporal locality when accessing the memory. Thus, the cache hit ratio can be guaranteed. A KNN query algorithm is given based on the MPMCQ framework and the grid index for moving objects. Extensive experiments are carried out to verify that by adopting the multi-threading and the cache optimization technologies, the proposed framework implements a much superior performance than other famous algorithms; moreover, it maintains excellent performance scalability when executed under different multi-core CPUs.

Abstract:Reprinting websites and blogs produces a great deal redundant WebPages. To improve search efficiency and user satisfaction, the near-Duplicate WebPages Detection based on Concept and Semantic network (DWDCS) is proposed. In the course of developing a near-duplicate detection system for a multi-billion pages repository, this paper makes two research contributions. First, the key concept is extracted, instead of the keyphrase, to build Small Word Network (SWN). This not only reduces the complexity of the semantic network, but also resolves the “expression difference” problem. Second, this paper considers both syntactic and semantic information to present and compute the documents’ similarities. In a large-scale test, experimental results demonstrate that this approach outperforms that of both I-Match and keyphrase extraction algorithms based on SWN. Many advantages such as linear time and space complexity, without using a corpus, make the algorithm valuable in actual practice.

Abstract:This paper proposes an efficient clustering method for protein sequences, using Affinity propagation algorithm (AP) and post-processing. In order to optimize the clustering result, post-processing is used to improve the clustering result of AP. To measure the similarity between two protein sequences, an improved alignment-free similarity measure is presented. This method is evaluated and compared with other algorithms on six protein sequences data sets. Experimental results demonstrate the effective performance of the proposed method.

Abstract:This paper analyzes the regular expression matching methods’ time complexity, space complexity and the tradeoff between them. The experiences, problems, and challenges encountered by the regular expression matching in network security field are well-classified and discussed in depth. Focusing on the two issues, a comprehensive overview of the current optimizing techniques and strategies adopted by academic and business communities is presented. Finally, a conclusion and some suggestions for future research are put forward.

Abstract:This paper gives the models for cross-layer mapping from services at the application-layer to multiple connections at the transport-layer then to multiple paths at the network-layer based on network utility maximization (NUM). This paper also presents the objective of cross-layer mapping from services to paths via connections, which is to allocate the path capacity of users, so that the aggregated utility of users can be globally maximized. For the mapping model, in order to achieve the optimum, a distributed algorithm is proposed which is asymptotically stable, and the equilibrium point is optimum. Simulation results verify the convergence of the proposed algorithm. Security and reliability of the mapping are also analyzed theoretically. The probabilities with which a service can be successfully completed are obtained when there are interceptions and distributed attacks in networks. Simulation examples are finally given to show the high security and reliability of the multi-to-multi mapping.

Abstract:To protect the spatial vector data, which is often in an irregular shape and distributed throughout multiple sensitive areas, the traditional mandatory access control model is extended and explained in this paper. This paper also proposes a fine-grained spatial mandatory query access control model?SV_MAC (spatial vector data mandatory access control model). Also, an AR+ spatial index tree technique is advanced, which combines the search of both spatial data and access control policies together to efficiently enforce the SV_MAC model in the course of spatial vector data searching. Experiment results shows that AR+ tree can not only provide fine-grained security protection for sensitive spatial vector data, but can also guarantee good user experience for GIS (geography information system) applications.

Abstract:This paper creates a kind of linear independent character code, presents a new theory based on Synergetic, which applies Synergetic to digital fingerprinting, and then proposes a scheme of collusion-resistant digital fingerprinting based on residual characters tracking. The efficiency and the performance are proved and analyzed by theories and examples. The experimental results indicate that the proposed method can greatly improve the efficiencies of encoding and tracing in the digital fingerprinting. The performance in robustness and collusion-resistant is able to meet the requirements of the applications of digital fingerprinting.

Abstract:In recent years, the detection technology based on machine learning algorithms for distributed denialof- service (DDoS) attacks has made great progress. However, there are still some deficiencies, which are: (1) being unable to make full use of contextual information in both the label and observed features series; (2) making too strong assumptions on the probability distribution of multiple features. Featured with the strong capability in integrating and exploiting contextual information and multiple features, the conditional random fields (CRF) model can be applied to detect DDoS attacks for effectively overcoming the above mentioned problems. A detection approach based on CRF model is proposed in this paper. First, two group of statistics are defined, which include traffic feature conditional entropy (TFCE) and behavior profile deviate degree (BPDD), to depict the characteristics of three types DDoS attacks: TCP flood, UDP flood and ICMP flood. Then, the CRF is trained to build the classification model for the addressed three types of attacks respectively. Lastly, the trained CRF models are used to identify the attacks with model inference. The experimental results demonstrate that the proposed approach can sufficiently exploit the advantages of CRF. The proposed detection approach not only can distinguish between attack traffic and normal traffic accurately, but is also more robust to resist disturbance of background traffic than the similar approaches.

Abstract:This paper reevaluates the security of Zodiac against impossible differential and integral attacks. In the past, results have shown that there are 15-round impossible differentials and 8-round integral distinguishers of Zodiac. Based on an 8-round truncated differential, with probability being 1, full 16-round impossible differentials and 9-round integral distinguishers are constructed. Integral attacks are applied to 12/13/14/15/16-round Zodiac with time complexities being 2³⁴, 2⁵⁹, 2⁹³, 2¹³³ and 2¹⁹⁰, respectively. Both the numbers of chosen plaintexts are no more than 2¹⁶, which shows that the full 16-round Zodiac is not immune to integral attack.

Abstract:Only six certificateless signcryption schemes have been proposed in recent years. Most of them cannot provide confidentiality and authentication. Even if some of them are secure, all of them need pairing operations. In order to solve the above-mentioned problems, a pairing-free certificateless signcryption scheme is proposed, and its security has proven to be in the random oracle model (ROM) under the computational Diffie-Hellman (CDH) assumption and the hardness of discrete logarithm problem (DLP). This scheme eliminates pairing operations and is the most efficient certificateless signcryption scheme.

Abstract:This paper presents a highly efficient algorithm for efficient order-independent transparency via compute unified device architecture (CUDA) in a single geometry pass. The study designs a CUDA renderer system to rasterize the scene by the scan-line algorithm, generating multiple fragments for each pixel. Meanwhile, a fixed size array is allocated per pixel in a GPU (graphics processing unit) global memory for storage. Next, this paper describes two schemes to capture and sorts the fragments per pixel via the atomic operations in CUDA. The first scheme stores the depth values of the fragments into an array of the corresponding pixel and sorts them on the fly using the atomicMin operation in CUDA. A following CUDA kernel will blend the fragments per pixel in depth order. The second scheme captures the fragments in rasterization order using the atomicInc operation in CUDA. During post-processing, the fragments per pixel array will be sorted in depth order before blending. Experimental result shows that this algorithm shows a significant improvement in classical depth peeling, producing faithful results.

Abstract:This paper presents an approach to animate realistic interactions between tree branches and raindrops, based on physical theory. Tree branches and petioles are represented by a model namely ETPSM (extended three-prism spring model) with three-prism structures, which can be flexibly controlled by four kinds of spring systems. The branches and leaves are animated due to the bidirectional transference of kinetic energy between raindrops and the branch system. The interactions between them can be well simulated by an efficient technique, specially designed for liquid motion on non-rigid objects with hydrophilic surfaces. When the branches are impacted by a raindrop, they will vibrate and twist, and the raindrops will flow along the leaf vein, merge into larger ones, or hang on the leaf apex, and the branch will rebound and vibrate for a while after the raindrop falls off the leaf, finally into a rest place. Experimental results show that this approach can be used to efficiently and realistically simulate these interactions between tree branches and raindrops; meanwhile, it can efficiently and easily simulate the elastic deformation of rod under the influence of external forces, such as rotation and vibration.

Abstract:Current PRT (pre-computed radiance transfer) techniques are limited to having 3D static scene, or large low-frequency lights. In this paper, an all-frequency shadow rendering method for dynamic scenes is proposed. In the preprocessing phase, the blocking geometry is modeled as a set of spheres, according to complex 3D model. The lighting and BRDF (bidirectional reflectance distribution function) are projected onto the Harr wavelet basis. At runtime, with the advantage of different basis functions, the product of visibility vectors is computed for blocker spheres in the pixel basis, while the triple product integral of lighting, BRDF and visibility is computed in the Harr basis. CUDA (computed unified device architecture) is used to implement this method, which sufficiently utilizes the new features of GPU (graphics processing unit). Experiments show that the method greatly improves vision quality and satisfies real-time requirements.

Abstract:This paper presents a method to describe and extract the general feature structures based on the fuzzy theory. This paper first builds some fuzzy measure rules with three levels (the basic property Φ, the derived property γ , and the associated property Π ). With the aid of measure rules, the feature vectors are established for fuzzy membership calculation and the feature region extraction. The analysis demonstrates that this method can obtain the optimal division of the flows on the principle of the minimum square sum. Further, the experiments show that the extent of typical flow structures extracted by this method are more effective than existing methods. In addition, the transfer function can be more flexible in design to avoid the cluttering and occlusion problems, which must be solved when visualizing 3D flows.