Search

The increased dependence on artificial neural network (ANN) models leads to a key question - will the ANN models provide accurate and reliable predictions? However, this important issue has received little systematic study. Thus this paper makes general researches on verification and validation (V&V) of ANN models. Basic problems for V&V of ANN models are explicitly presented, a new V&V approach for ANN models is developed, V&V methods for ANN models are deeply discussed, further research areas for V&V of ANN models are recommended, and an example is given. [ABSTRACT FROM AUTHOR]

In utilities using a mixture of hydroelectric and non-hydroelectric power, the economics of the hydroelectric plants depend upon the reservoir height and the inflow into the reservoir for several months into the future. Accurate forecasts of reservoir inflow allow the utility to feed proper amounts of fuel to individual plants, and to economically allocate the load between various non-hydroelectric plants. For this reasons, several companies in the Brazilian Electrical Sector use the linear time-series models such as PARMA (Periodic Auto regressive Moving Average) models. This paper provides for river flow prediction a numerical comparison between constructive neural networks and PARMA models. The results obtained in the evaluation of the performance of Neural Network were better than the results obtained with PARMA models. [ABSTRACT FROM AUTHOR]

In this paper, a novel license plate location algorithm for color image is presented. Firstly the neural networks are used as filters for analyzing within small windows for an image and deciding whether each window contains a license plate or not coarsely. And then we use the information which the license plate's saturation value is different from the background's, so it can be used to locate license plate finely. At last, color pairs method is presented to prove whether the region we found is the license plate region or not. The experimental results show that proposed algorithms are robust in dealing with the license plate location in complex background. [ABSTRACT FROM AUTHOR]

Design and optimization of the flight controllers is a demanding task which usually requires deep engineering knowledge of intrinsic aircraft behavior. In this study, EAs are used to design a controller for recovery (landing) of a small fixed-wing UAV (Unmanned Aerial Vehicle) on a frigate ship deck. This paper presents an approach in which the whole structure of the control laws is evolved. The control laws are encoded in a way common for Genetic Programming. However, parameters are optimized independently using effective Evaluation Strategies, while structural changes occur at a slower rate. The fitness evaluation is made via test runs on a comprehensive 6 degree-of-freedom non-linear UAV model. The results show that an effective controller can be designed with little knowledge of the aircraft dynamics using appropriate evolutionary techniques. An evolved controller is demonstrated and a set of reliable algorithm parameters is identified. [ABSTRACT FROM AUTHOR]

The web logs contain the information of the user's access record to a web site. The recommender system of the web site is improved by analyzing web log file including user's duration time at each web page. The web usage mining is the application of data mining techniques to large web data repositories in order to extract usage patterns. Many algorithms have been proposed to construct recommender system in web usage mining. In general, the size of web log records is large. So we have difficulties to analyze web log data. To make matter worse, the web log data are very sparse. It is very hard to estimate the dependencies between the web pages. Therefore, we solved these problems of web usage mining using combined evolutionary computing into support vector machine. In this paper, we proposed a new mining model for web usage mining. We verified the performance of proposed model using two data sets from KDD Cup 2000 and our web server. [ABSTRACT FROM AUTHOR]

It is a big challenge to guarantee the quality of association rules in some application areas (e.g., in information gathering) since duplications and ambiguities of data values (terms). This paper presents a novel concept of rough association rules to improve the quality of discovered knowledge. The precondition of a rough association rule consists of a set of terms (items) and a weight distribution of terms (items). The distinct advantage of rough association rules is that they contain more specific information than normal association rules. [ABSTRACT FROM AUTHOR]

It is known that algorithms for discovering association rules generate an overwhelming number of those rules. While many new very efficient algorithms were recently proposed to allow the mining of extremely large datasets, the problem due to the sheer number of rules discovered still remains. In this paper we propose a new way of pushing the constraints in dual-mode based from the set of maximal patterns that is an order of magnitude smaller than the set of all frequent patterns. [ABSTRACT FROM AUTHOR]

Both automated design and knowledge discovery of electronic circuits are challenging tasks for artificial intelligence. A genetic algorithm (GA) based approach to them was proposed in this paper, which features an array-based encoding scheme, a multi-objective evaluation mechanism and an adaptation strategy for GA parameters. It was validated by the experiments on arithmetic circuits of gradually increasing scales, which evolved multi-objective optimized circuits and revealed some novel and generalized principles. [ABSTRACT FROM AUTHOR]

There have been many technical trading rules in stock market since the first stock exchange founded. Along with the developing of computer technology, the technical trading rules are playing more and more important roles in the stock market trading system. However, there are many problems also occurred, such as the huge database, inefficiency, etc. So, the in-depth data mining technology is becoming a powerful tool to overcome the shortage of the current technologies. In this paper, we give some applications of in-depth data mining method: to find the optimal range, to find the stock-rule pair and find the relationship between the number of pair and investment. This method can improve both efficiency and effectiveness. [ABSTRACT FROM AUTHOR]

In this paper, we report on the use of ant systems in the data mining field capable of extracting comprehensible classifiers from data. The ant system used is a - ant system which differs from the originally proposed ant systems in its ability to explore bigger parts of the solution space, yielding better performing rules. Furthermore, we are able to include intervals in the rules resulting in less and shorter rules. Our experiments show a significant improvement of the performance both in accuracy and comprehensibility, compared to previous data mining techniques based on ant systems and other state-of-the-art classification techniques. [ABSTRACT FROM AUTHOR]

Mobile phones are becoming a popular platform for a range of applications. However, due to size restrictions, the interfaces of these applications can be difficult to use. Customising an interface for a particular user offers the potential to improve an interface's efficiency. In this paper, we propose customising a mobile phone's Profile application. We apply a machine learning approach to discover concepts that describe a user's profile-activations in terms of their scheduled appointments. We found that it is possible to learn useful concepts, which maybe used to improve the users interaction with mobile phone devices. [ABSTRACT FROM AUTHOR]

This paper proposes the model based detection method of abnormal acoustic source using a microphone array. General source location algorithm using a microphone array can be used to locate a dominant acoustic source, while this does not verify whether the detected source is permitted one or not on outdoor environments. It is difficult to discern it among a natural environmental sound. Thus, to cope with this problem, we propose the out-of-normal acoustic rejection method based on N-best likelihood ratio test using natural environmental sound models. In order to evaluate the proposed algorithm, a real-time DSP was constructed, and experimental evaluation is described. [ABSTRACT FROM AUTHOR]

Creating Petri Net domain models faces the same challenges that confront all knowledge-intensive AI performance systems: model specification, knowledge acquisition, and refinement. Thus, a fundamental question to investigate is the degree to which automation can be used. This paper formulates the learning task and presents the first machine learning method for Time Interval Petri Net (TIPN) domain models. In a preliminary evaluation within a damage control domain, the method learned a nearly perfect model of fire spread augmented with temporal and spatial data. Keywords: domain model learning, Petri net learning, spatial-temporal data series learning, real-time decision-making, automated damage control. [ABSTRACT FROM AUTHOR]

To minimize the outliers' effects, in this paper, a new speaker identification scheme based on the t-distribution mixture model is proposed. Since the t-distribution provides a longer and heavier tailed alternative to the Gaussian distribution, the mixture model with multivariate t-distribution is expected to show more robust results than the Gaussian mixture model(GMM) in the cases where outliers exist. In experiments, we compared the performance of the proposed scheme with that of using the conventional GMM to show its robustness. [ABSTRACT FROM AUTHOR]

Most classification algorithms suffer from an inability to detect instances of classes which are not present in the training set. A novel approach for characteristic concept rule learning called IC2 is proposed in this paper. [ABSTRACT FROM AUTHOR]

Clustering analysis is used to analyze the clustering phenomenon occurred to the data structure. However, there are some problems when the decision maker attempts to use clustering analysis. For solving these existing problems, this paper proposes a revised Entropy Clustering Analysis method requiring no prior setting of clusters, which is based on the mean distance between the data points and the cluster center. Through using several experiments and comparing different clustering analysis methods with proposed method, the results show that the proposed clustering method could achieve reasonable clustering effect. The experiment also proves that using the attributes with high correlation coefficient in clustering can achieve higher clustering accuracies. [ABSTRACT FROM AUTHOR]

Inspired by the theory of multi-resolution analysis of wavelet transform, combining advantages of multi-resolution theory and support vector machine, a new regression model that is called multi-resolution support vector regression (MR-SVR) for function regression is proposed in this paper. In order to construct MR-SVR, the scaling function at some scale and wavelets with different resolution is used as kernel of support vector machine, which is called multi-resolution kernel. The MR-SVR not only has the advantages of support vector machine, but also has the capability of multi-resolution which is useful to approximate nonlinear function. Simulation examples show the feasibility and effectiveness of the method. [ABSTRACT FROM AUTHOR]

This paper presents a novel hybrid clustering approach that takes advantage of the efficiency of k-Means clustering and the effectiveness of hierarchical clustering. It employs the combination of geometrical information defined by k-Means and topological information formed by the Voronoi diagram to advantage. Our proposed approach is able to identify clusters of arbitrary shapes and clusters of different densities in O(n) time. Experimental results confirm the effectiveness and efficiency of our approach. [ABSTRACT FROM AUTHOR]

A novel Feature-Watermarking algorithm based on wavelet packet decomposition was presented in this paper. We first propose the concept of Feature-Watermark. Dither modulation embedding scheme in wavelet packet coefficients promises the hiding of large capacity of robust information and fulfills watermark blind-extraction. Experimental results show that our method successfully fulfills the compromise between the robustness and capacity. [ABSTRACT FROM AUTHOR]

This paper deals with relationship between BP learning for neural networks and numerical algorithm of differential equations. It is proposed that the iteration formula of BP algorithm is equivalent to Euler method of differential dynamic system under certain conditions, and the asymptotic solutions of the two formulas are consistent. It is also proved in theoretic that asymptotic solutions given by BP algorithm are equivalent to that computed by any numerical method for differential dynamic systems under certain conditions. Also, an example to train the BP network by modified numerical method is presented. [ABSTRACT FROM AUTHOR]

Estimation of distribution algorithms (EDAs) are popular and robust algorithms that combine two technical disciplines of soft computing methodologies, probabilistic reasoning and evolutionary computing, for optimization problems. Several algorithms have already been proposed by different authors. However, these algorithms may require huge computation power, which is seldom considered in those applications. This paper introduces a "fast estimation of distribution algorithm" (FEDA) for feature selection that does not evaluate all new individuals by actual fitness function, thus reducing the computational cost and improve the performance. Bayesian networks are used to model the probabilistic distribution and generate new individuals in the optimization process. Moreover, fitness value is assigned to each new individual using the extended Bayesian network as an approximate model to fitness function. Implementation issues such as individual control strategy, model management are addressed. Promising results are achieved in experiments on 5 UCI datasets. The results indicate that, as population-sizing requirements for building appropriate models of promising solutions lead to good fitness estimates, more compact feature subsets that give more accurate result can be found. [ABSTRACT FROM AUTHOR]

For classification of health data, we propose in this paper a fast and accurate feature selection method, FIEBIT (Feature Inclusion and Exclusion Based on Information Theory). FIEBIT selects the most relevant and non-redundant features using Conditional Mutual Information (CMU) while excluding irrelevant and redundant features according to the comparison among Individual Symmetrical Uncertainty (ISU) and Combined Symmetrical Uncertainty (CSU). Small feature subsets are selected before classification without compromising the classification accuracy. In addition, the size of the feature subset is determined automatically. Our preliminary empirical results on health data with hundreds of features suggest FIEBIT is efficient and effective in comparison with representative feature selection methods. [ABSTRACT FROM AUTHOR]

Inducing sequential patterns from time series data is an important data mining problem. While most of the current methods are generating sequential patterns within a single attribute, this paper proposes a new method, using Hellinger entropy measure, for generating multi-dimensional sequential patterns. A number of theorems are proposed to reduce the computational complexity of the proposed method. [ABSTRACT FROM AUTHOR]

Recently, there are several nonlinear dimensionality reduction algorithms that can discover the low-dimensional coordinates on a manifold based on training samples, such as ISOMAP, LLE, Laplacian eigenmaps. However, most of these algorithms work in batch mode. In this paper, we presented an incremental nonlinear dimensionality reduction algorithm to efficiently map new samples into the embedded space. The method permits one to select some landmark points and to only preserve geodesic distances between new data and landmark points. Self-organizing map algorithm is used to choose landmark points. Experiments demonstrate that the proposed algorithm is effective. [ABSTRACT FROM AUTHOR]

Text information processing depends critically on the proper text representation. A common and naïve way of representing a document is a bag of its component words [1], but the semantic relations between words are ignored, such as synonymy and hypernymy-hyponymy between nouns. This paper presents a model for representing a document in terms of the synonymy sets (synsets) in WordNet [2]. The synsets stand for concepts corresponding to the words of the document. The Vector Space Model describes a document as orthogonal term vectors. We replace terms with concepts to build Concept Vector Space Model (CVSM) for the training set. Our experiments on the Reuters Corpus Volume I (RCV1) dataset have shown that the result is satisfactory. Keywords: Data mining, ontology, knowledge discovery. [ABSTRACT FROM AUTHOR]

With mixed feature data, problems are induced in modeling the gating network of normalized Gaussian (NG) networks as the assumption of multivariate Gaussian becomes invalid. In this paper, we propose an independence model to handle mixed feature data within the framework of NG networks. The method is illustrated using a real example of breast cancer data. [ABSTRACT FROM AUTHOR]

The objective of this paper is to propose an intelligent security management system using simulation based analysis, which is capable to monitor network status, evaluate vulnerabilities, generate defense strategies, and apply it to the network. To do this, we have employed the intelligent system design concept based on the advanced modeling and simulation environment for developing network security models and simulation-based evaluation of vulnerability as well as defense strategy. Our approach differs from others in that i) it is able to analyze vulnerabilities on node, link, and network in quantitative manner, ii) it can generate and apply defense strategies automatically, and iii) it supports a coherent design concept for intelligent security management system. A case study performed on a test bed will illustrate our techniques and demonstrate effectiveness of propose system. [ABSTRACT FROM AUTHOR]

Multiple sequence alignment (MSA) is a fundamental and challenging problem in the analysis of biologic sequences. In this paper, an immune particle swarm optimization (IPSO) is proposed, which is based on the models of the vaccination and the receptor editing in immune systems. The proposed algorithm is used to train hidden Markov models (HMMs), further, an integration algorithm based on the HMM and IPSO for the MSA is constructed. The approach is tested on a set of standard instances taken from the Benchmark Alignment database, BAliBASE. Numerical simulated results are compared with those obtained by using the Baum-Welch training algorithm. The results show that the proposed algorithm not only improves the alignment abilities, but also reduces the time cost. [ABSTRACT FROM AUTHOR]

This paper investigates the problem of obstacle avoidance and path planning for robotic fish. The swimming of the robot fish to avoid some obstacles is viewed as potential flow around the obstacles. Then the streamlines from the robot position to the target are chosen as the desired paths for the mobile robot to move to the destination. Since there are mature algorithms with high computational efficiency to establish flow field and figure out the streamlines based on fluid mechanics theory, our approach is practical for application. We conduct two example experiments to verify the effectiveness of the approach. [ABSTRACT FROM AUTHOR]

Globalisation of business environments on the Web has given rise to the advent of similar ontologies in which dynamic ontology mapping is called in. Ontology mapping is necessary for ontology interoperability. In this paper, a novel agent-based ontology mapping is presented for agents to operate ontology mapping flexibly in a dynamic environment regardless heterogeneous platforms and different ontology representations. The mapping mechanism is discussed by having a close look at both inherent inter-processes of mapping tasks of an agent and relevant interaction processes. The interrelated processes of agents also enable agent-based ontology mapping to take ontology changes into account whenever needed. A mapping prototype is built for verification. [ABSTRACT FROM AUTHOR]

The optimization of job-shop scheduling is very important because of its theoretical and practical significance. This paper proposes an efficient scheduling method based on artificial immune systems. In the proposed method, the initial population is generated by a proposed scheduling initialization algorithm based on the G&T algorithm, and the models of the vaccination and receptor editing are designed to improve the immune performance. The approach is tested on a set of standard instances taken from the existing standard library. The simulation results validate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

In this paper we present several efficient computational procedures for defeasible reasoning while the plausible and well-defined semantics, viz.preferred models and stable models, are not given up. The proposed algorithms exploit the structural information of defeasible knowledge bases to facilitate efficient computational models. [ABSTRACT FROM AUTHOR]

This paper describes a method for unfolding a Predicate-net representation of a logistical planning problem, such that it possesses the Markov property. The problem can then be easily converted into a Markov Decision Process (MDP) which can be solved in a tractable manner using standard Dynamic Programming algorithms. [ABSTRACT FROM AUTHOR]

In this paper, a basic architecture of implemented multi-agent ship design system is introduced briefly. And then several cases of conflicts occurred in designing process are described. Finally, conflict resolution method based on case-based reasoning (CBR) approach is presented. Through the help of the developed multi-agent ship design system, a designer can make decisions or resolve some conflicts easily based on the previous resolved similar cases. [ABSTRACT FROM AUTHOR]

Knowledge reduction is one of the key problems of rough set theory. In this paper, we investigate some theoretical issues of the reduction of information systems and present a new reduction approach. A closure operator on the power set of attributes is defined. The relations between the closed sets determined by the closure operator and the indiscernibility attribute sets are then investigated. Based on the relations, we can determine a partition on the power of attributes by using the indiscernibility attribute sets. Consequently, the reducts of any subset of attributes in information systems can be derived. [ABSTRACT FROM AUTHOR]

Plausible Logic is a non-monotonic logic with an efficient implementation, but no semantics. This paper gives Plausible Logic a fixed-point semantics, similar to the extensions of Reiter's Default Logic. The proof theory is sound but deliberately incomplete with respect to this semantics. This is because the semantics is an attempt to define what follows from a plausible theory, rather than merely giving a different characterisation of what is provable. [ABSTRACT FROM AUTHOR]

Plausible Logic is a non-monotonic logic with an efficient implementation. Plausible Logic has five proof algorithms, one is monotonic and four are non-monotonic. These five proof algorithms form a hierarchy. Ambiguity propagating proof algorithms are less risky than ambiguity blocking proof algorithms. The hierarchy shows that the benefit of using the riskier algorithms is that more formulas can be proved. Unlike previous Plausible Logics, the Plausible Logic in this paper is relatively consistent, checks for loops, can prove all its facts and all tautologies, and allows countably many formulas and rules to be considered. [ABSTRACT FROM AUTHOR]

This paper considers the use of randomly generated directed graphs as neighborhoods for particle swarm optimizers (PSO) using fully informed particles (FIPS), together with dynamic changes to the graph during an algorithm run as a diversity-preserving measure. Different graph sizes, constructed with a uniform out-degree were studied with regard to their effect on the performance of the PSO on optimization problems. Comparisons were made with a static random method, as well as with several canonical PSO and FIPS methods. The results indicate that under appropriate parameter settings, the use of random directed graphs with a probabilistic disruptive re-structuring of the graph produces the best results on the test functions considered. [ABSTRACT FROM AUTHOR]

Multi-class object classification is an important field of research in computer vision. In this paper basic linear genetic programming is modified to be more suitable for multi-class classification and its performance is then compared to tree-based genetic programming. The directed acyclic graph nature of linear genetic programming is exploited. The existing fitness function is modified to more accurately approximate the true feature space. The results show that the new linear genetic programming approach outperforms the basic tree-based genetic programming approach on all the tasks investigated here and that the new fitness function leads to better and more consistent results. The genetic programs evolved by the new linear genetic programming system are also more comprehensible than those evolved by the tree-based system. [ABSTRACT FROM AUTHOR]

Shadow detection and removal is important to deal with traffic image sequences. The shadow cast by a vehicle can lead to inaccurate object feature extraction and an erroneous scene analysis. Furthermore, separate vehicles can be connected through a shadow, thereby confusing an object recognition system. Accordingly, this paper proposes a robust method for detecting and removing an active cast shadow from monocular color image sequences. A background subtraction method is used to extract moving blobs in color and gradient dimensions, and YCrCb color space adopted to detect and remove the cast shadow. Even when shadows link different vehicles, each vehicle figure can be separately detected using a modified mask based on a shadow bar. Experimental results from town scenes demonstrate that the proposed method is effective and the classification accuracy is sufficient for general vehicle type classification. [ABSTRACT FROM AUTHOR]

Hierarchical structure, reusable and dynamic components, and predictable interactions are distinct characteristics of hybrid intelligent systems (HIS). The existing agent-oriented methodologies are deficient in HIS construction because they did not take into account the characteristics of HIS. In this paper, we propose a Methodology for constructing Agent-based HIS (MAHIS). MAHIS consists of eight models: Hybrid Strategy Identification Model, Organization Model, Task Model, Agent Model, Expertise Model, Coordination Model, Reorganization Model, and Design Model. The Reorganization Model is the key model to support dynamic platform-based HIS. It consists of category role, group roles, virtual organization role, and dynamics rules. This model describes the characteristics of HIS with virtual organization, category, and group perspectives. Some previously developed agents can be reused by means of involving them in a new virtual organization dynamically. The output of the Reorganization Model is the specification of the dynamic platform which comprises middle agents and makes all agents and agent groups hierarchical and dynamic. [ABSTRACT FROM AUTHOR]

As social entities, intelligent agents need to collaborate with others, regardless of whether they are cooperative or self-interested. The durations of agent collaborations can be long-term or "one-shot". Nowadays, many multi-agent system applications require the system to work in open and dynamic domains. In such dynamic environments, how long collaboration should be kept among particular agents are always a problem to be discussed. In this paper, we focus on general self-interested multi-agent systems and analyze the advantages and disadvantages that can be brought by one-shot teams and long-term teams. Furthermore, we present a mechanism that can enable agents to form teams with reasonable terms and objects. Keywords: Multi-agent system, team formation, self-interested agent, dynamic team forming. [ABSTRACT FROM AUTHOR]

In a multi-agent environment, there is often the need for an agent to cooperate with others so as to ensure that a given task is achieved timely and cost-effectively. Present agent systems currently maximizes this through mechanisms such as trust and risk assessments. In this paper, we extend this mechanism by introducing the concept of insurance, in which the insurance agents act as a bridge between agents who require resources from others. Unlike traditional systems, agents purchase insurance so as to guarantee to have the requested resources during the task execution time and thus minimize the risk in task failure. The novelty of this proposal is that it ensures agents continuously to exchange resources and to seek maximum expected utility in a dynamic environment at the same time. Our experimental results confirm the feasibility of our approach. [ABSTRACT FROM AUTHOR]

The paper proposes new approach for modelling negotiation partners and predictive decision-making. It is based on the prediction of negotiation partners behaviour from its previous offers in the current encounter. The approach allows the negotiating agent to asses different factors influencing other agent's behaviour during negotiation and make optimal decisions according to the prediction. It is tested in simple scenario and the results illustrating the comparison with random strategy selection are presented. [ABSTRACT FROM AUTHOR]

In this paper we proposed a framework for collaborative intelligent agents in a distributed environment to execute sound security strategies to protect information resources. First, the intelligent agent based Duty Reliable Center in the framework uses group decision method to determine a global information threat level. Then local agent employs the threat level and applies the Bayes' decision procedure to calculate expected loss of its all possible actions, and choices an action among them with minimum expected loss to protect its information resources. The model enables an agent to choose among alternatives in an optimal fashion, taking into account the worth of acquiring prior information to reduce the uncertainty. Because system operations are distributed, hackers are unlikely to wreck the whole system. Thus, it is expected to yield information security cost-effective solutions. [ABSTRACT FROM AUTHOR]

All of the Prototype Reduction Schemes (PRS) which have been reported in the literature, process time-invariant data to yield a subset of prototypes that are useful in nearest-neighbor-like classification. In this paper, we suggest two time-varying PRS mechanisms which, in turn, are suitable for two distinct models of non-stationarity. In both of these models, rather than process all the data as a whole set using a PRS, we propose that the information gleaned from a previous PRS computation be enhanced to yield the prototypes for the current data set, and this enhancement is accomplished using a LVQ3-type "fine tuning". The experimental results, which to our knowledge are the first reported results applicable for PRS schemes suitable for non-stationary data, are, in our opinion, very impressive. Keywords: Prototype Reduction Schemes (PRS), Time Varying Samples (TVS), Nonstatinoary Environments, Hybrid-type Prototype Reduction. [ABSTRACT FROM AUTHOR]

Text documents have sparse data spaces, and nearest neighbors may belong to different classes when using current existing proximity measures to describe the correlation of documents. In this paper, we propose an asymmetric similarity measure to strengthen the discriminative feature of document objects. We construct a semantic correlation network by asymmetric similarity between documents and conjecture the power law feature of the connections distributions. Hub points which exist in semantic correlation network are classified by an agglomerative hierarchical clustering approach named SCN. Both objects similarity and neighbors similarity are considered in the definition of hub points proximity. Finally, we assign the rest text objects to their nearest hub points. The experimental evaluation on textual data sets demonstrates the validity and efficiency of SCN. The comparison with other clustering algorithms shows the superiority of our approach. Keywords: Data Mining, Knowledge Discovery. [ABSTRACT FROM AUTHOR]

This paper presents a complete visual surveillance system for automatic scene interpretation of airport aprons. The system comprises two main modules — Scene Tracking and Scene Understanding. The Scene Tracking module is responsible for detecting, tracking and classifying the semantic objects within the scene using computer vision. The Scene Understanding module performs high level interpretation of the observed objects by detecting video events using cognitive vision techniques based on spatio-temporal reasoning. The performance of the system is evaluated for a series of pre-defined video events specified using a video event ontology. [ABSTRACT FROM AUTHOR]

Eigenspace models are a convenient way to represent set of images with widespread applications. In the traditional approach to calculate these eigenspace models, known as batch PCA method, model must capture all the images needed to build the internal representation. This approach has some drawbacks. Since the entire set of images is necessary, it is impossible to make the model build an internal representation while exploring a new object. Updating of the existing eigenspace is only possible when all the images must be kept in order to update the eigenspace, requiring a lot of storage capability. In this paper we propose a method that allows for incremental eigenspace update method by incremental kernel PCA for vision learning and recognition. Experimental results indicate that accuracy performance of proposed method is comparable to batch KPCA and outperform than APEX. Furthermore proposed method has efficiency in memory requirement compared to KPCA. Content Area: Vision, robotics. [ABSTRACT FROM AUTHOR]

This paper presents a new method that improves the performance of the speaker specific phonetic template based speech recognizer with the speaker-independent (SI) phoneme HMMs. Since the phonetic template based speech recognizer uses only the phoneme transcription of the input utterance, the performance of the system is worse than that of the speaker dependent system due to the mismatch between the training data and the SI models. In order to solve these problems, a new training method that iteratively estimates the phonetic templates and transformation vectors for the adaptation of the SI phoneme HMMs is presented. The phonetic class based and codebook-based stochastic matching methods are used to estimate the transformation vectors for speaker adaptation. Performance evaluation using the speaker dependent recognition experiments performed over actual telephone line showed a reduction of about 40% in the error rates when compare to the conventional speaker specific phonetic template based speech recognizer. [ABSTRACT FROM AUTHOR]