International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
1
6
New Adaptive Method to Speed Up Training of Back Propagation Algorithm
Zahraa Elsayed Mohamed , Elsadek Hussien Ibrahim
Zagazig University, Faculty of Science, Zagazig, Egypt
High Institute of Engineering and Technology Al-Obour, Zagazig, Egypt
ABSTRACT
Using Back Propagation (BP) algorithm in training of multilayer feedforward networks is a very effective learning approach. It finds the best weights of Artificial Neural Network (ANN) by computing the weight change. There are many drawbacks of BP algorithm but the main problems are slow training and reaching local minima easily. Through the last years, a lot of algorithms were proposed to improve and modify the BP algorithm. Overcoming these problems requires adding new parameters as learning rate and momentum. In this research, a new adaptive BP algorithm is proposed by introducing a new function for adaptive learning rate and adaptive momentum which depends on the error gradient at every layer. In the learning samples, the simulation results mention the convergence action of proposed algorithm. Comparing with classic BP algorithm, the proposed algorithm gives better convergence rates and finds a good solution efficiently. Three popular benchmark classification problems are used to explain the improvement in convergence rates.
Code
MPH
STA
NAM
DSA
CSC
ALG
TGA
FUA
DEA
page
01
25
43
55
61
67
79
90
102
International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
2
6
Performance Analysis of Some Ad hoc Routing Protocols AODV, DSR and OLSR under Certain Conditions as Speed, Network Density and Packet Size
Zahraa E. Mohamed , Mahmoud Atef
Faculty of Science, Mathematics Department, Zagazig University, Zagazig, Egypt
High Institute for Engineering & Technology – Al-Obour, Cairo, Egypt
ABSTRACT
MANET is a collection of mobile nodes which are independent, self-organization and self-configuration where each node communicates with other nodes by a multi-hops manner and can move arbitrarily without any infrastructure or central control. This paper aims to analysis and compare the performance of three ad hoc routing protocols namely: AODV, DSR and OLSR vs. three different variables parameters using OPNET 17.5 simulations. Furthermore, this study focuses on comparing the main features of these protocols and evaluated the performance of these protocols. We studied the effect of nodes density, mobility speed and data packet size on the performance of these protocols based on the rate of file transfer protocol (FTP) with medium load traffic. The results in all simulations decided that the performance of OLSR protocol is better than both of them AODV and DSR in terms of end to end delay and data dropped, Whereas the performance AODV protocol is the best in terms of throughput and also DSR protocol has the lowest network load. As, we concluded that each protocol has different effect with respect to the environment conditions and considered metrics, including the average throughput, network load, end-to-end average delay, and data dropped. The results are shown that OLSR protocol can be more suitable choice in a large density networks compared to AODV and DSR.
Code
MPH
STA
NAM
DSA
CSC
ALG
TGA
FUA
DEA
page
01
25
43
55
61
67
79
90
102
International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
3
6
Implementation of the particle swarm optimization algorithm for solving the JSSP
using parallel MATLAB
Wael M.F. Abdel-Rehim
Faculty of computers and information, Suez University, Egypt.
ABSTRACT
In this study, we implement a Particle Swarm Optimization (PSO)-based method in parallel by using a parallel MATLAB with one, two, three, and four threads to solve the Job-Shop Scheduling Problem (JSSP). The resulting parallel PSO algorithm is evaluated by applying it to some job shop benchmark problems. The obtained results indicate that implementing PSO in parallel is an effective method for the JSSP that significantly increases the speedups especially for large-scale problems.
Code
MPH
STA
NAM
DSA
CSC
ALG
TGA
FUA
DEA
page
01
25
43
55
61
67
79
90
102
International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
4
6
Binary versus non-binary visual features representations: a comparative study
Hammam A. Alshazly , M. Hassaballah , Abdelmgeid Amin Ali
Faculty of Science, South Valley University, Qena, Egypt
Faculty of Computers and Information, South Valley University, Luxor, Egypt
Faculty of Computers and Information, Minia University, El Minia, Egypt
ABSTRACT
Many computer vision applications rely on detecting a set of salient image features, then representing the neighborhood of each feature point by a feature vector for further processing. In this paper, we introduce a comparative study between the two main categories of representing the feature neighborhood: binary and non-binary. We highlight their differences under various image conditions using image datasets that have geometric and photometric transformations and different scene types. In addition, we illustrate when to pick a specific representation scheme based on the application needs or the distortions exist in the images.
Code
MPH
STA
NAM
DSA
CSC
ALG
TGA
FUA
DEA
page
01
25
43
55
61
67
79
90
102
International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
5
6
Self-Stabilization Deterministic Cluster-Based Skip List Protocol for Peer to Peer Systems
Ahmed A. A. Gad-ElRab , T. A. A. Alzohairy , Khaled A. A. Khalf-Allah
Department of Mathematics, Faculty of Science Al-Azhar University - Cairo, Egypt
Department of Mathematics, Faculty of Science Al-Azhar University - Cairo, Egypt
Department of Mathematics, Faculty of Science Al-Azhar University - Cairo, Egypt
ABSTRACT
Self-stabilization is one of the most important concepts used to build an overlay network which is a logical layer that used to manage and manipulate the information and data of a dynamic distributed systems as Peer-to-Peer (P2P) system. In P2P system the topology of the network is continuously changing by joining and leaving peers at any time. Stabilization ensures that P2P system is a persistently running system with a countless number of peers which communicate with each other and every peer has a partial view of the network. In this paper, the algorithms of Deterministic Cluster-Based Skip List protocol for dynamic distributed system (CBSL) stability will be analyzed and proved by using self-stabilization methods such as linearization, correctness and connectivity, and the ability to extend CBSL to a ring
Code
MPH
STA
NAM
DSA
CSC
ALG
TGA
FUA
DEA
page
01
25
43
55
61
67
79
90
102
International Conference on Mathematics, Trends and Development ICMTD17 ,
Cairo, Egypt, 28 – 30 Dec. 2017 Organized by The Egyptian Mathematical Society,
Web Site: http://www.etms-eg.org
Computer Science.
6
6
Predicting and Analysis of Students' Performance Using data mining Techniques
Reda Mohammed , Nahla F. Omran , Abdelmgeid A. Ali
Faculty of Science, South Valley University, Qena, Egypt
Faculty of Science, South Valley University, Qena, Egypt
Faculty of Computers and Information, Minia University, Al Minia, Egypt
ABSTRACT
Educational database holds on massive amount of data and it is increasing rapidly. Data mining provides effective techniques for discovering useful knowledge and pattern from students' data. The discovered patterns can be used to understand many problems in the educational field. This paper proposes a framework to predict the performance of first year bachelor’s students in computer science course. Decision Tree, Naïve Bayes, and Multilayer Perceptron classification methods are applied to the students’ data using the Weka Data Mining tool to produce the best prediction model of the students’ academic performance. We conduct experiments that detect the best model among the used techniques and compute models' accuracy. The extracted knowledge from prediction model will be utilized to recognize and profile the student to decide the students' level of success in the first semester.
