فهرست مطالب

Information Systems and Telecommunication - Volume:6 Issue: 3, 2018 Jul-Sep

Journal of Information Systems and Telecommunication
Volume:6 Issue: 3, 2018 Jul-Sep

  • تاریخ انتشار: 1398/01/17
  • تعداد عناوین: 7
|
  • Hassan Hafez, Kolahi, Shohreh Kasaei* Pages 119-127
    Information Theory (IT) has been used in Machine Learning (ML) from early days of this field. In the last decade, advances in Deep Neural Networks (DNNs) have led to surprising improvements in many applications of ML. The result has been a paradigm shift in the community toward revisiting previous ideas and applications in this new framework. Ideas from IT are no exceptions. The fast rate of new publications, the diversity of seemingly unrelated applications, and the longtime-span of previous ideas which are being revisited, make a challenge for a researcher to view the whole picture. In this survey, these problems are mostly addressed by giving an organized review on the vast amount of recent publications on the intersection of IT and ML, while mentioning their connections to previous ideas. The focus is on Information Bottleneck (IB), a formulation of information extraction based on IT which is recently found to be a good candidate in studying DNNs. Two important concepts are highlighted in this narrative on the subject, (i) the concise and universal view that IT provides on seemingly unrelated methods of ML, demonstrated by explaining how IB relates to minimal sufficient statistics, stochastic gradient descent, and variational auto-encoders, and (ii) the common technical mistakes and problems caused by applying ideas from IT, which is discussed by a careful study of some recent methods suffering from them.
    Keywords: Machine Learning, Information Theory, Information Bottleneck, Deep Learning, Variational Auto- Encoder
  • Hanieh Zamanian, Hassan Farsi*, Sajad Mohamadzadeh Pages 128-135
    Recent researches on pixel-wise semantic segmentation, use deep neural networks to improve accuracy and speed of these networks in order to increase the efficiency in practical applications such as automatic driving. These approaches have used deep architecture to predict pixel tags, but the results seem to be undesirable. The reason for these unacceptable results is mainly due to the existence of max pooling operators, which reduces the resolution of the feature maps. In this paper, we present a convolutional neural network composed of encoder-decoder segments based on successful SegNet network. The encoder section has a depth of 2, which in the first part has 5 convolutional layers, in which each layer has 64 filters with dimensions of 3×3. In the decoding section, the dimensions of the decoding filters are adjusted according to the convolutions used at each step of the encoding. So, at each step, 64 filters with the size of 3×3 are used for coding, the weights of these filters are adjusted by network training and adapted to the educational data. Due to having the low depth of 2, and the low number of parameters in the proposed network, the speed and the accuracy improve in comparison with the popular networks such as SegNet and DeepLab. For the CamVid dataset, after a total of 60,000 iterations, we attain the 91% for global accuracy, which indicates improvements in the efficiency of the proposed method
    Keywords: semantic segmentation, convolutional neural networks, encoder –decoder, pixelwise semantic interpretation
  • Hamed Agahi*, Azar Mahmoodzadeh, Marzieh Salehi Pages 136-149
    Handwritten digit recognition, as an important issue in pattern classification, has received considerable attention of many researchers to develop theoretical and practical aspects of this problem. The goal is to recognize a printed digit text or a scanned handwritten using an automated procedure. In this paper an optical character recognition system with a multi-step procedure is presented for Farsi handwritten digit classification. First, a pre-processing course is performed on the image to enhance and make prepare the image for the main steps. Then multiple features are extracted which are believed to be effective in the classification step, among which a multi-objective genetic algorithm selects those with more discriminative characteristics in order to reduce the computational complexity of the classification steps. Following this, only the selected features are extracted from the digit images to be entered to three classifiers. Once the classifiers are trained, their outputs are fed into a classifier ensemble to make the final decision. The weights of the linear combination of classifiers are adjusted by an evolutionary firefly algorithm which optimizes the F-criterion. Evaluation of the proposed technique on the standard HODA database demonstrates that the algorithm of this paper attains higher performance indices including F-measure of 98.88%, compared to other existing methods.
    Keywords: Classifiers ensemble, Feature extraction, Feature selection, Firefly algorithm, Multi-objective genetic algorithm, Optical character recognition
  • Faisal Aburub* Pages 150-158
    A number of different hotels have been seen to direct significant investment towards ERP systems with the aim of securing sound levels of organizational learning. As a strategic instrument, organizational learning has been recommended in the modern management arena as potentially able to achieve a competitive edge and as stabilizing the success of businesses. Learning, as an aim, is not only able to improve the skillset and knowledge of employees, but also achieving organizational growth and development, whilst also helping to build a dynamic learning organization. Organizational learning is especially important in modern-day firms, when staff might choose to leave or change their role owing to the view that knowledge-sharing could be detrimental to their own success. The present work seeks to examine the impact of ERP usage on organizational learning. A new research model has been presented, this model has been empirically investigated in the Jordanian hotel industry. The findings from the empirical findings emphasize that ERP usage has significant impact on organizational learning. In line with the study findings, various aspects of organizational learning, such as continuous learning, system perspective, openness and experimentation and transfer and integration are recognized as able to best encourage the use of ERP. Suggestions for future work and discussion on research limitations are also discussed.
    Keywords: ERP Usage, Organizational Learning, Organizational Performance
  • Farzad Bayat, Zahra Nilforoushan* Pages 159-168
    Tree search algorithms are vital for the search methods in structured data. The core of such algorithms is nodes, which can be taken from a data structure. One famous tree data structure is split tree. In this paper, to compute the split tree in polar coordinates, a method has been introduced. Assuming that the algorithm inputs (in form of points) have been distributed in the form of a circle or part of a circle, polar split tree can be used. For instance, we can use these types of trees to transmit radio and telecommunication waves from host stations to the receivers and to search the receivers. Since we are dealing with the figures which are approximately circular, it is better to use polar coordinates. Furthermore there are several researches by search algorithms for the central anchor which leads to the assignment of a virtual polar coordinate system. In this paper, firstly the structure of Cartesian split tree will be explained and then the polar split tree will be implemented. Eventually, by doing nearest neighbor search experiments, we will compare the polar split tree and polar quad tree in terms of searching time and amount of distance to the closest neighbor and in the end, better results will be achieved.
    Keywords: split tree, polar split tree, quad tree, polar quad tree, nearest neighbor search
  • Amir Hossein Hosseinian, Vahid Baradaran* Pages 169-179
    This paper addresses the community detection problem as one of the most significant problems in the field of social network analysis. The goal of the community detection problem is to find sub-graphs of a network where they have high density of within-group connections, while they have a lower density of between-group connections. Due to high practical usage of community detection in scientific fields, many researchers developed different algorithms to meet various scientific requirements. However, single-objective optimization algorithms may fail to detect high quality communities of complex networks. In this paper, a novel multi-objective multi-agent optimization algorithm, named the MAOA is proposed to detect communities of complex networks. The MAOA aims to optimize modularity and community score as objective functions, simultaneously. In the proposed algorithm, each feasible solution is considered as an agent and the MAOA organizes agents in multiple groups. The MAOA uses new search operators based on social, autonomous and self-learning behaviors of agents. Moreover, the MAOA uses the weighted sum method (WSM) in finding the global best agent and leader agent of each group. The Pareto solutions obtained by the MAOA is evaluated in terms of several performance measures. The results of the proposed method are compared with the outputs of three meta-heuristics. Experiments results based on five real-world networks show that the MAOA is more efficient in finding better communities than other methods.
    Keywords: Community detection problem, Complex networks, Multi-agent systems, Social networks
  • Omid Mahdi Ebadati E*, Farshad Eshghi, Amin Zamani Pages 180-192
    Wireless sensor networks are new technologies that are used for various purposes such as environmental monitoring, home security, industrial process monitoring, healthcare programs and etc. The involvement of wireless communication technology also acquires different types of security threats. Wireless sensor networks are vulnerable to various attacks. Cryptography is one of the methods for secure transmission of information between sensors in wireless sensor networks. Security can be established in each layer of application, network, data link, and physical layer. In an encryption scheme, the information or message set to be transmitted using an encryption algorithm change so, that it can only be read by decrypting. A complete and secure encryption system must establish three principles of confidentiality, authentication and integrity. An encryption algorithm alone cannot provide all the principles of encryption. A hybrid encryption algorithm, consisting of symmetric and asymmetric encryption algorithms, provides complete security for a cryptographic system. The papers presented in this area over the last few years, and a new secure algorithm present with regard to the limitations of wireless sensor networks, which establishes three principles of cryptography. The details of the algorithm and basic concepts are presented in such a way that the algorithm can be operational and showed a very high efficiency in compare to the current proposed methods.
    Keywords: Wireless sensor network, Cryptography algorithm, Hybrid cryptography, Confidentiality Integration Authentication