Earticle

The interests of web users can be revealed by their visited web pages and time duration on these web pages during their surfing. Time duration on a web page is characterized as a fuzzy linguistic variable because linguistic variable makes users easily understand the expression of time duration and can disregard subtle difference between two time durations. Each web access pattern from web logs is transformed as corresponding fuzzy web access pattern, which is a fuzzy vector composed of fuzzy linguistic variables or 0. Each element in fuzzy web access patterns represents visited web page and time duration on this web page. This paper proposed a rough k-means clustering algorithm based on properties of rough variable to group the gained fuzzy web access patterns. Finally, an example and experiment is provided to illustrate the clustering process. Using this approach, users can effectively mine web logs records to discover interesting user access patterns.

Multi-process bidirectional heuristic search algorithms (such as PBA*) that utilize island nodes have been shown to have the potential for exponential speedup over their plain counterparts that do not utilize island nodes. PBA* (Parallel Bidirectional A*) is a primary algorithm in this area and, to the best of our knowledge, unique in its usage of island nodes for reducing the size of the traversed search space and in its usage of multidimensional heuristics for reducing interprocess communication cost. Both, island nodes use and reference nodes use within the context of multidimensional heuristics have been reported to provide excellent performance, compared to plain heuristic search counterparts. However, both of these two features have also resisted refinements for over two decades. Specifically, the two main issues are: how to identify well-placed (in the state space) island nodes, and how to generate well-placed reference nodes. The work presented in this paper is an initial step toward this end. We present two methods, one employing Voronoi diagrams for the purpose of identifying well-placed island nodes, and one employing Kohonen Networks for the purpose of identifying well-placed reference nodes. We implement and compare our methods against a plain version of PBA*. Our finding indicate that the cost of incorporating the proposed improvements into PBA* is negligible; and when PBA* is equipped with the island identification method, or with the reference node identification method, it outperforms its random island nodes counterpart and its random reference nodes counterpart, respectively, for the vast majority of test cases.

In this paper we investigates time delay and channel gain estimation for multipath fading Code Division Multiple Access (CDMA) signals using the second order Divided Difference Filter (DDF). Given the nonlinear dependency of the channel parameters on the received signals in multiuser/multipath scenarios, we show that the DDF achieves better performance than its linear counterparts. The DDF, which is a derivative-free Kalman filtering approach, avoids the errors associated with linearization in the conventional Extended Kalman Filter (EKF). The Numerical results also show that the proposed DDF is simpler to implement, and more resilient to near-far interference in CDMA networks.

Artificial K-lines (AKL) is a structure that can be used to store different types of knowledge, as long as this knowledge is represented by series of events connected by causality. Unlike, and, perhaps, complementary to, Artificial Neural Networks (ANN), AKL can combine inter-domain knowledge and its knowledge base can be augmented dynamically without rebuilding of the entire system. In this paper we demonstrate the diversity of AKL by illustrating, through examples, its workings for three applications across three completely different areas of study. The first example demonstrates that our structure can generate a solution where most other known technologies are either incapable of, or very complicated in doing so. The second example illustrates a novel, human-like, way of machine learning. The third example presents a behavior metrics based method for password authentication.

Security is one of the most important issues in E-government. All of the security approaches that are common in E-commerce are applicable to E-government. But Egovernment is a little different from E-commerce. Usually government networks can communicate to each other better than business networks, because, most of them are connected for transferring information, but businesses are competitors and they don't disclose their sensitive information. Utilizing "honeypots" is a good solution for tracing hackers and revealing their tools. In this paper, "connectedness" of E-government networks and honeypots are employed to propose an approach for securing an E-government network. This framework provides suitable resources for hackers; and simultaneously, it prevents them from misusing those resources for future attacks.

Delta modulation plays a key role in data communication; the problem encountered in delta modulation is the slope over load error, which is inherent in the system. In order for the signal to have good fidelity, the slope-overload error needs to be as small as possible. Adaptive delta modulation reduces the slope over load error to a greater extent. ADM attempts to increase the dynamic range and the tracking capabilities of fixed step-size delta modulation. The adaptive algorithms adjust the step size (from a range of step sizes) to the power level of the signal and thus enhance the dynamic range of the coding system appreciably. This paper suggests a novel 1-bit Adaptive Delta Modulation technique for improving the signal-to-noise ratio (SNR) of Adaptive Delta Modulators (ADM). Various step-size algorithms are discussed and also their performance comparison is made. A new technique has also been discussed with a suitable pre-filer used for improving the SNR.