Recent Patents on Computer Science - Volume 11, Issue 2, 2018

Minimum-margin nearest neighbor instance selection is one of the data reduction techniques for the feed-forward neural network that can improve the scalability and make incremental learning computationally feasible. The technique utilizes the Euclidean distance function to produce a reduced training set that can conserve the distribution model of the original training set and yield a comparable level of the classification accuracy of the neural network as the original training set. Nevertheless, the technique does not consider the range of the attributes, causing some attributes to dominate the other attributes on the Euclidean distance calculation. This paper studies an integration of six normalization techniques; Min-Max, Z-score, Mean MAD, Median MAD, Hyperbolic Tangent Estimator, and Modified Hyperbolic Tangent, with the minimum-margin nearest neighbor instance selection algorithms to improve the data reduction performance, the execution time reduction performance of the data reduction techniques, and the classification performance of the feedforward neural network. The experimental results on the real-world datasets from the UCI database and ELENA project confirmed that the Min-Max, Z-score, Mean MAD, and Median MAD normalization techniques could improve the data reduction performance more efficient than the Modified Hyperbolic Tangent and Hyperbolic Tangent Estimator normalization techniques. The Median MAD normalization could improve the execution time reduction performance of the data reduction techniques most efficiently. Besides, most normalization techniques could improve the accuracy, error rate, precision, recall, and F1-score of the feed-forward neural network when the final training set included both the normalized original training sets and the reduced normalized training sets.

Background: Regardless of the significance of medium scale search engines, a few scholarly researches are based on them. Besides, we regularly hear about innovation and web expansion in this research area. Recently, making a web search engine is altogether different from the past years. Here, we introduce MSSE (Medium Scale Search Engine). Methods: Recent publications and patent databases of search engines are reviewed. MSSE is a prototype of a medium scale search engine. It makes overwhelming utilization of the existing structure in hypertext. MSSE is intended to crawl and index the Web content efficiently. The search in this engine produces accepting results competing for conventional systems. Results: The proposed model of this work was introduced as prototype. The prototype introduced was at least of 500000 pages. These pages were associated with a full text (content) and hyperlink database. In addition, we use OPNET modeler as a metrics environment of MSSE in order to verify the obtained results about the performance of the search engine. Conclusion: We provided an in-depth description of MSSE. This work addresses this inquiry of how to design and create a practical medium scale system that can add extra information introduce in hypertext.

Background: Parsing English language achieved effective results over the last few decades. However, parsing a difficult language such as Arabic represents a major challenge at the present, since it is characterized by the rich morphology and contains complex linguistic characteristics not found in other languages. Although parsing systems for Arabic have been developed recently, however, most of them do not support any deeper processing for the Arabic sentences such as providing an effective dependencies analysis to identify, for example, the relative clauses in these sentences. Objective: This paper develops a new framework and system that support the process of parsing Arabic sentences and writing well-formed Arabic relative clauses. Method: The developed framework is applied to learn the grammar rules for Arabic relative clauses based on the use of machine learning, in particular, Inductive Logic Programming (ILP). A corpus of Arabic relative sentences was generated from Quran and used in the experiments made in this research. The sentences in this corpus were firstly processed by using the Natural Language Processing (NLP) toolkit called Stanford coreNLP and then given to the ILP system ALEPH to automatically learn a grammar for Arabic relative clauses. A system was developed to extract Arabic relative clauses from Arabic sentences based on the rules produced by ALEPH. Results: An empirical evaluation of the developed system was carried out and achieved promising results with an overall accuracy of 83%. Conclusion: Our results lead to conclude that the developed system is able to perform a deeper dependency parsing for Arabic text as well as it can identify relative clauses in Arabic sentences.

Background: The main objective of the proposed mechanism is to minimize the energy consumption of the entire network and to maximise the MANET lifetime by balancing the energy level of the node. Selection of angle based high battery nodes is necessary since the nodes with higher energy drain rate present in the path and the nodes moving away from the path lead to network partition. Therefore, preserving the node from high drain rate is essential to maintain the better link quality. Here, power aware routing protocol is used to determine path discovery with low routing requests as well as minimises the total transmission power required to reach the destination. Conclusion: Stable routes are determined by applying this proposed technique which avoids the nodes with less battery power during route discovery phase. The intermediate nodes present in the stable routes have high residual capacity of power. Also, the transmission power control algorithm is proposed to identify high transmission data rate nodes present in the transmission range towards the destination. The Energy Efficient Data Forwarder (EEDF) nodes can be preserved with high energy and power. Simulation analysis for the proposed method is carried out using NS-2 tool for performance evaluation.

Background: Scarcity in frequency availability for communication systems prompts a dire need to explore technologies to avail existing available spectrum much more efficiently. Cognitive Radio Network (CRN) technology is one such promising solution. This technology permits the unlicensed users or Secondary Users (SU) to utilize the spectrum allocated to licensed users (Primary Users), when not in use by Primary Users (PU). But, SUs must do spectrum handoff whenever a PU wants its channel again for its own use. Also, the SUs continuously sense the channels to identify an idle channel. Error in the sensing channel is possible. Method: Here, spectrum sensing errors are analyzed for non-uniform unlicensed users, in cognitive radio ad-hoc network. A detection theory is put forth, to sense error. In addition, it is possible that in a network, SUs may not have the same amount of data packets to transmit. Result: It leads to variation in the number of frames i.e. non-uniform distribution. This effect is studied and a practical approach is proposed to calculate throughput. Average normalized SU throughput is compared with a signal to noise ratio using random and pseudorandom channel selection algorithm. Conclusion: Thus, the simulation results are obtained and the effects are analyzed in the performance of overall network. The proposed model can be duly made use of to more practical network scenarios.