SEARCH RESULT

Neural networks (NNs) is used to solve many engineering and science problem. Generally, feedforward architecture is preferred and gradient-based learning algorithms are extensively operated to tune all parameters of NN iteratively. This training method is a conventional one, but training process takes a long time due to the slowness of gradient-based learning algorithms. This slowness has been an important drawback in their applications. To overcome this disadvantage, extreme learning machine (ELM) concept introduced to science community in near past. Essentially, ELM is a data-driven learning algorithm for single-hidden layer feedforward neural networks (SLFNs). This algorithm provides extremely fast learning speed. In this study, performance of SLFNs learned by ELM algorithm is investigated on the problem of highly nonlinear dynamic system identification. As a result of studies on selected benchmark problems in the literature, it has been seen that ELM may not provide a good generalization success due to randomly chosen the number of hidden nodes and weight parameters for inputs in SLFN. For both the training and the test data set, very poor results have been obtained and observed surprisingly during the above-mentioned studies. Here, a simple heuristic approach has been proposed in this study in order to eliminate this bad situation and the findings obtained with this approach are discussed. Based on the obtained experimental results, it has been shown that the proposed approach determines the optimal the number of hidden nodes and a reasonable random selection of input weights required for a good generalization performance.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Cihan Karakuzu Uğur Yüzgeç

RC4 is an algorithm that encrypts a data in the form of a bit string with a specified key. The security of RC4 with high encryption speed depends on the random key. Speed of image encryption is very important parameter due to size of the data. Performance analysis was examined using color image encryption, because RC4 is usually used in speed-critical applications. RC4 showed that the desired values could not be obtained when examining the histogram, correlation coefficient and information entropy analysis results. Because of this, chaotic systems are used to increase the performance criteria of image encryption with RC4. Chaotic systems are very sensitive due to their inherent dependence on the initial conditions and dynamic variables. Not random and non-periodic oscillations these systems are performed in a certain frequency range. The RC4 algorithm are enhanced by using chaotic system-based encryption algorithms because of its key size capability and speed. Performance criteria have been improved using 2D Cat Map, Tent Map and Lorenz chaotic systems. When we examine only the encryption made with RC4 and the encryption made with RC4 supported with chaotic systems; more successful results are obtained from histogram, correlation coefficient and knowledge entropy analysis. In addition, the structure of chaotic systems is increased key sensitivity and key size and thus a more secure algorithm is obtained.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Sefa Tunçer Cihan Karakuzu F. UÇAR

In this study, the Opposition based learning Spiral Optimization Algorithm (OBLSOA) is presented for training Multi-Layer Perceptron (MLP). The main idea of Spiral Optimization Algorithm (SOA) is based on the dynamic step dimension in its spiral path trajectory. The primary opposition based learning (OBL) concept first was come from the Yin-Yang symbol in the ancient Chinese philosophy. According to OBL concept, if a candidate point is far from the solution, the opposite point of this candidate can be closer to the solution than that point. We applied OBL concept to spiral optimization algorithm for training MLP. OBLSOA comprises two main stages: the first is the opposition-based learning population initialization and the other is opposition-based learning generation jumping. To evaluate the performance of the proposed OBLSOA, we used eight standard datasets including four classification datasets (XOR, balloon, Iris, breast cancer) and three function-approximation datasets (sigmoid, cosine, and sine). The performance proposed OBLSOA was compared with the original SOA for all datasets in terms of the Mean Square Error (MSE) metric. The training and test results show that the proposed OBLSOA is able to be provide very competitive and effective in training MLPs.

International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES

Uğur Yüzgeç Cihan Karakuzu