Teknik Bilimler MYO,Bilgisayar Programcılığı, Bildiri & Sunum Koleksiyonu
Bu koleksiyon için kalıcı URI
Güncel Gönderiler
Öğe A model of deep neural network for iris classification with different activation functions(IEEE, 2018) Eldem, Ayşe; Eldem, Hüseyin; Üstün, DenizIn recent years, deep neural network (DNN) has been frequently used for classification. In this study, iris flowers having 3 different types are classified by using DNN which are utilized the width and length of petal and sepal features as input. Some experiments are made for the iris dataset by using different activation functions and different epoch numbers. Then, the activation function which gives the best result is determined. The classification success of the developed model is achieved 96% for the iris dataset.Öğe An improved opportunistic packet transmission scheme for wireless sensor networks(IEEE, 2010) Soy, Hakkı; Özdemir, Özgür; Bayrak, MehmetIn this paper, we consider a wireless sensor network in a single cell architecture, where a single hop between sensor nodes and central controller node exists. We propose a new medium access control (MAC) protocol and packet transmission scheme between sensor nodes with single antenna and central controller node with multiple antennas. Our scheme is an extension to previous work proposed by Coronel et al. [In Proceedings of the IEEE International Conference on Communications, Vol. 2, pp. 1082-1086, 2005] for wireless sensor networks. Unlike previous approach, enabling the use of the multi-user diversity based on normalized signal to noise ratio, we aim to increase throughput and energy efficiency in the system by increasing the received packets and reducing collisions. The throughput performance of the suggested scheme is compared through Monte Carlo simulations for frequency flat Rayleigh slow fading channel model.Öğe Throughput analysis of SNR and NSNR based opportunistic wireless sensor networks(2011) Soy, Hakkı; Özdemir, Ö.; Bayrak, MehmetWe previously proposed a threshold based medium access control (MAC) protocol that can perform multiuser diversity gain for star topology based wireless sensor network in a single cell architecture. The considered network model comprises a single hop between the sensor nodes and the controller node with multiple antennas. The optimum threshold was designed such that at a given time the probability that only one of the sensor nodes is above the threshold is being maximum. In this paper, the throughput performance of this MAC protocol is analyzed theoretically that has been based on the signal to noise ratio (SNR) and the normalized SNR (NSNR) estimation at the sensor nodes having a single antenna. Furthermore, the throughput and fairness performances of these are measured through Monte Carlo simulations for Rayleigh slow fading channel model. The obtained results shown that the tradeoff between the throughput and fairness can be adjusted by channel estimation metric..
