Learning Flexible Neural Networks for Pattern Recognition

Learning the gradient of neuron's activity function like the weight of links causes a new specification which is flexibility. In flexible neural networks because of supervising and controlling the operation of neurons, all the burden of the learning is not dedicated to the weight of links, therefore in each period of learning of each neuron, in fact the gradient of their activity function, cooperate in order to achieve the goal of learning thus the number of learning will be decreased considerably. Furthermore, learning neurons parameters immunes them against changing in their inputs and factors which cause such changing. Likewise initial selecting of weights, type of activity function, selecting the initial gradient of activity function and selecting a fixed amount which is multiplied by gradient of error to calculate the weight changes and gradient of activity function, has a direct affect in convergence of network for learning.

The Performance of the Character-Access on the Checking Phase in String Searching Algorithms

A new algorithm called Character-Comparison to Character-Access (CCCA) is developed to test the effect of both: 1) converting character-comparison and number-comparison into character-access and 2) the starting point of checking on the performance of the checking operation in string searching. An experiment is performed; the results are compared with five algorithms, namely, Naive, BM, Inf_Suf_Pref, Raita, and Circle. With the CCCA algorithm, the results suggest that the evaluation criteria of the average number of comparisons are improved up to 74.0%. Furthermore, the results suggest that the clock time required by the other algorithms is improved in range from 28% to 68% by the new CCCA algorithm

Progressive AAM Based Robust Face Alignment

AAM has been successfully applied to face alignment, but its performance is very sensitive to initial values. In case the initial values are a little far distant from the global optimum values, there exists a pretty good possibility that AAM-based face alignment may converge to a local minimum. In this paper, we propose a progressive AAM-based face alignment algorithm which first finds the feature parameter vector fitting the inner facial feature points of the face and later localize the feature points of the whole face using the first information. The proposed progressive AAM-based face alignment algorithm utilizes the fact that the feature points of the inner part of the face are less variant and less affected by the background surrounding the face than those of the outer part (like the chin contour). The proposed algorithm consists of two stages: modeling and relation derivation stage and fitting stage. Modeling and relation derivation stage first needs to construct two AAM models: the inner face AAM model and the whole face AAM model and then derive relation matrix between the inner face AAM parameter vector and the whole face AAM model parameter vector. In the fitting stage, the proposed algorithm aligns face progressively through two phases. In the first phase, the proposed algorithm will find the feature parameter vector fitting the inner facial AAM model into a new input face image, and then in the second phase it localizes the whole facial feature points of the new input face image based on the whole face AAM model using the initial parameter vector estimated from using the inner feature parameter vector obtained in the first phase and the relation matrix obtained in the first stage. Through experiments, it is verified that the proposed progressive AAM-based face alignment algorithm is more robust with respect to pose, illumination, and face background than the conventional basic AAM-based face alignment algorithm.

Network Anomaly Detection using Soft Computing

One main drawback of intrusion detection system is the inability of detecting new attacks which do not have known signatures. In this paper we discuss an intrusion detection method that proposes independent component analysis (ICA) based feature selection heuristics and using rough fuzzy for clustering data. ICA is to separate these independent components (ICs) from the monitored variables. Rough set has to decrease the amount of data and get rid of redundancy and Fuzzy methods allow objects to belong to several clusters simultaneously, with different degrees of membership. Our approach allows us to recognize not only known attacks but also to detect activity that may be the result of a new, unknown attack. The experimental results on Knowledge Discovery and Data Mining- (KDDCup 1999) dataset.

Optimizing Mobile Agents Migration Based on Decision Tree Learning

Mobile agents are a powerful approach to develop distributed systems since they migrate to hosts on which they have the resources to execute individual tasks. In a dynamic environment like a peer-to-peer network, Agents have to be generated frequently and dispatched to the network. Thus they will certainly consume a certain amount of bandwidth of each link in the network if there are too many agents migration through one or several links at the same time, they will introduce too much transferring overhead to the links eventually, these links will be busy and indirectly block the network traffic, therefore, there is a need of developing routing algorithms that consider about traffic load. In this paper we seek to create cooperation between a probabilistic manner according to the quality measure of the network traffic situation and the agent's migration decision making to the next hop based on decision tree learning algorithms.

LINUX Cluster Possibilities in 3-D PHOTO Quality Imaging and Animation

In this paper we present the PC cluster built at R.V. College of Engineering (with great help from the Department of Computer Science and Electrical Engineering). The structure of the cluster is described and the performance is evaluated by rendering of complex 3D Persistence of Vision (POV) images by the Ray-Tracing algorithm. Here, we propose an unexampled method to render such images, distributedly on a low cost scalable.

A Perceptual Image Coding method of High Compression Rate

In the framework of the image compression by Wavelet Transforms, we propose a perceptual method by incorporating Human Visual System (HVS) characteristics in the quantization stage. Indeed, human eyes haven-t an equal sensitivity across the frequency bandwidth. Therefore, the clarity of the reconstructed images can be improved by weighting the quantization according to the Contrast Sensitivity Function (CSF). The visual artifact at low bit rate is minimized. To evaluate our method, we use the Peak Signal to Noise Ratio (PSNR) and a new evaluating criteria witch takes into account visual criteria. The experimental results illustrate that our technique shows improvement on image quality at the same compression ratio.

Recursive Algorithms for Image Segmentation Based on a Discriminant Criterion

In this study, a new criterion for determining the number of classes an image should be segmented is proposed. This criterion is based on discriminant analysis for measuring the separability among the segmented classes of pixels. Based on the new discriminant criterion, two algorithms for recursively segmenting the image into determined number of classes are proposed. The proposed methods can automatically and correctly segment objects with various illuminations into separated images for further processing. Experiments on the extraction of text strings from complex document images demonstrate the effectiveness of the proposed methods.1

A Content Based Image Watermarking Scheme Resilient to Geometric Attacks

Multimedia security is an incredibly significant area of concern. The paper aims to discuss a robust image watermarking scheme, which can withstand geometric attacks. The source image is initially moment normalized in order to make it withstand geometric attacks. The moment normalized image is wavelet transformed. The first level wavelet transformed image is segmented into blocks if size 8x8. The product of mean and standard and standard deviation of each block is computed. The second level wavelet transformed image is divided into 8x8 blocks. The product of block mean and the standard deviation are computed. The difference between products in the two levels forms the watermark. The watermark is inserted by modulating the coefficients of the mid frequencies. The modulated image is inverse wavelet transformed and inverse moment normalized to generate the watermarked image. The watermarked image is now ready for transmission. The proposed scheme can be used to validate identification cards and financial instruments. The performance of this scheme has been evaluated using a set of parameters. Experimental results show the effectiveness of this scheme.

Learning FCM by Tabu Search

Fuzzy Cognitive Maps (FCMs) is a causal graph, which shows the relations between essential components in complex systems. Experts who are familiar with the system components and their relations can generate a related FCM. There is a big gap when human experts cannot produce FCM or even there is no expert to produce the related FCM. Therefore, a new mechanism must be used to bridge this gap. In this paper, a novel learning method is proposed to construct causal graph based on historical data and by using metaheuristic such Tabu Search (TS). The efficiency of the proposed method is shown via comparison of its results of some numerical examples with those of some other methods.

Contour Estimation in Synthetic and Real Weld Defect Images based on Maximum Likelihood

This paper describes a novel method for automatic estimation of the contours of weld defect in radiography images. Generally, the contour detection is the first operation which we apply in the visual recognition system. Our approach can be described as a region based maximum likelihood formulation of parametric deformable contours. This formulation provides robustness against the poor image quality, and allows simultaneous estimation of the contour parameters together with other parameters of the model. Implementation is performed by a deterministic iterative algorithm with minimal user intervention. Results testify for the very good performance of the approach especially in synthetic weld defect images.

Dynamic Clustering using Particle Swarm Optimization with Application in Unsupervised Image Classification

A new dynamic clustering approach (DCPSO), based on Particle Swarm Optimization, is proposed. This approach is applied to unsupervised image classification. The proposed approach automatically determines the "optimum" number of clusters and simultaneously clusters the data set with minimal user interference. The algorithm starts by partitioning the data set into a relatively large number of clusters to reduce the effects of initial conditions. Using binary particle swarm optimization the "best" number of clusters is selected. The centers of the chosen clusters is then refined via the Kmeans clustering algorithm. The experiments conducted show that the proposed approach generally found the "optimum" number of clusters on the tested images.

Class Outliers Mining: Distance-Based Approach

In large datasets, identifying exceptional or rare cases with respect to a group of similar cases is considered very significant problem. The traditional problem (Outlier Mining) is to find exception or rare cases in a dataset irrespective of the class label of these cases, they are considered rare events with respect to the whole dataset. In this research, we pose the problem that is Class Outliers Mining and a method to find out those outliers. The general definition of this problem is “given a set of observations with class labels, find those that arouse suspicions, taking into account the class labels". We introduce a novel definition of Outlier that is Class Outlier, and propose the Class Outlier Factor (COF) which measures the degree of being a Class Outlier for a data object. Our work includes a proposal of a new algorithm towards mining of the Class Outliers, presenting experimental results applied on various domains of real world datasets and finally a comparison study with other related methods is performed.

WiPoD Wireless Positioning System based on 802.11 WLAN Infrastructure

This paper describes WiPoD (Wireless Position Detector) which is a pure software based location determination and tracking (positioning) system. It uses empirical signal strength measurements from different wireless access points for mobile user positioning. It is designed to determine the location of users having 802.11 enabled mobile devices in an 802.11 WLAN infrastructure and track them in real time. WiPoD is the first main module in our LBS (Location Based Services) framework. We tested K-Nearest Neighbor and Triangulation algorithms to estimate the position of a mobile user. We also give the analysis results of these algorithms for real time operations. In this paper, we propose a supportable, i.e. understandable, maintainable, scalable and portable wireless positioning system architecture for an LBS framework. The WiPoD software has a multithreaded structure and was designed and implemented with paying attention to supportability features and real-time constraints and using object oriented design principles. We also describe the real-time software design issues of a wireless positioning system which will be part of an LBS framework.

Data Annotation Models and Annotation Query Language

This paper presents data annotation models at five levels of granularity (database, relation, column, tuple, and cell) of relational data to address the problem of unsuitability of most relational databases to express annotations. These models do not require any structural and schematic changes to the underlying database. These models are also flexible, extensible, customizable, database-neutral, and platform-independent. This paper also presents an SQL-like query language, named Annotation Query Language (AnQL), to query annotation documents. AnQL is simple to understand and exploits the already-existent wide knowledge and skill set of SQL.

Robot Task-Level Programming Language and Simulation

This paper presents the development of a software application for Off-line robot task programming and simulation. Such application is designed to assist in robot task planning and to direct manipulator motion on sensor based programmed motion. The concept of the designed programming application is to use the power of the knowledge base for task accumulation. In support of the programming means, an interactive graphical simulation for manipulator kinematics was also developed and integrated into the application as the complimentary factor to the robot programming media. The simulation provides the designer with useful, inexpensive, off-line tools for retain and testing robotics work cells and automated assembly lines for various industrial applications.

On the Reduction of Side Effects in Tomography

As the Computed Tomography(CT) requires normally hundreds of projections to reconstruct the image, patients are exposed to more X-ray energy, which may cause side effects such as cancer. Even when the variability of the particles in the object is very less, Computed Tomography requires many projections for good quality reconstruction. In this paper, less variability of the particles in an object has been exploited to obtain good quality reconstruction. Though the reconstructed image and the original image have same projections, in general, they need not be the same. In addition to projections, if a priori information about the image is known, it is possible to obtain good quality reconstructed image. In this paper, it has been shown by experimental results why conventional algorithms fail to reconstruct from a few projections, and an efficient polynomial time algorithm has been given to reconstruct a bi-level image from its projections along row and column, and a known sub image of unknown image with smoothness constraints by reducing the reconstruction problem to integral max flow problem. This paper also discusses the necessary and sufficient conditions for uniqueness and extension of 2D-bi-level image reconstruction to 3D-bi-level image reconstruction.

Estimation of Buffer Size of Internet Gateway Server via G/M/1 Queuing Model

How to efficiently assign system resource to route the Client demand by Gateway servers is a tricky predicament. In this paper, we tender an enhanced proposal for autonomous recital of Gateway servers under highly vibrant traffic loads. We devise a methodology to calculate Queue Length and Waiting Time utilizing Gateway Server information to reduce response time variance in presence of bursty traffic. The most widespread contemplation is performance, because Gateway Servers must offer cost-effective and high-availability services in the elongated period, thus they have to be scaled to meet the expected load. Performance measurements can be the base for performance modeling and prediction. With the help of performance models, the performance metrics (like buffer estimation, waiting time) can be determined at the development process. This paper describes the possible queue models those can be applied in the estimation of queue length to estimate the final value of the memory size. Both simulation and experimental studies using synthesized workloads and analysis of real-world Gateway Servers demonstrate the effectiveness of the proposed system.

Extraction of Significant Phrases from Text

Prospective readers can quickly determine whether a document is relevant to their information need if the significant phrases (or keyphrases) in this document are provided. Although keyphrases are useful, not many documents have keyphrases assigned to them, and manually assigning keyphrases to existing documents is costly. Therefore, there is a need for automatic keyphrase extraction. This paper introduces a new domain independent keyphrase extraction algorithm. The algorithm approaches the problem of keyphrase extraction as a classification task, and uses a combination of statistical and computational linguistics techniques, a new set of attributes, and a new machine learning method to distinguish keyphrases from non-keyphrases. The experiments indicate that this algorithm performs better than other keyphrase extraction tools and that it significantly outperforms Microsoft Word 2000-s AutoSummarize feature. The domain independence of this algorithm has also been confirmed in our experiments.

A Dynamic Time-Lagged Correlation based Method to Learn Multi-Time Delay Gene Networks

A gene network gives the knowledge of the regulatory relationships among the genes. Each gene has its activators and inhibitors that regulate its expression positively and negatively respectively. Genes themselves are believed to act as activators and inhibitors of other genes. They can even activate one set of genes and inhibit another set. Identifying gene networks is one of the most crucial and challenging problems in Bioinformatics. Most work done so far either assumes that there is no time delay in gene regulation or there is a constant time delay. We here propose a Dynamic Time- Lagged Correlation Based Method (DTCBM) to learn the gene networks, which uses time-lagged correlation to find the potential gene interactions, and then uses a post-processing stage to remove false gene interactions to common parents, and finally uses dynamic correlation thresholds for each gene to construct the gene network. DTCBM finds correlation between gene expression signals shifted in time, and therefore takes into consideration the multi time delay relationships among the genes. The implementation of our method is done in MATLAB and experimental results on Saccharomyces cerevisiae gene expression data and comparison with other methods indicate that it has a better performance.