Browsing by Author "Ozmen, Atilla"

Now showing 1 - 5 of 5

Citation - WoS: 1
Citation - Scopus: 1
Age Classification by WGAN Brain MR Image Augmentation
(IEEE, 2024) Yaman, Batuhan; Yilmaz, Ozge Zeynep; Darici, Muazzez Buket; Ozmen, Atilla
Medical image augmentation plays a crucial role in enhancing the performance of Artificial Intelligence (AI) applications in medical sciences. Augmenting medical images is important for solving data scarcity, increasing data diversity, enhancing robustness and reliability of model and improving training and test results that can be done in medical sciences. In this work we show that Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) can be used for increasing the performance of data classification. To achieve that, we have augmented healthy brain MR images by using WGAN and updated the dataset. The results give that when dataset augmented by WGAN-GP is used as input for CNN-based model to solve age classification problem, accuracy of this model increases to 98,37% from 95,14%. It can be concluded that the purposed WGAN-based brain MR image augmentation method enhances the performance of image classification.
Citation - WoS: 89
Citation - Scopus: 120
Air Quality Prediction Using Cnn Plus Lstm-Based Hybrid Deep Learning Architecture
(Springer Heidelberg, 2022) Gilik, Aysenur; Ogrenci, Arif Selcuk; Ozmen, Atilla
Air pollution prediction based on variables in environmental monitoring data gains further importance with increasing concerns about climate change and the sustainability of cities. Modeling of the complex relationships between these variables by sophisticated methods in machine learning is a promising field. The objectives of this work are to develop a supervised model for the prediction of air pollution by using real sensor data and to transfer the model between cities. The combination of a convolutional neural network and a long short-term memory deep neural network model was proposed to predict the concentration of air pollutants in multiple locations of a city by using spatial-temporal relationships. Two approaches have been adopted: the univariate model contains the information of one pollutant whereas the multivariate model contains the information of all pollutants and meteorology data for prediction. The study was carried out for different pollutants which are in the publicly available data of the cities of Barcelona, Kocaeli, and Istanbul. The hyperparameters of the model (filter, frame, and batch sizes; number of convolutional/LSTM layers and hidden units; learning rate; and parameters for sample selection, pooling, and validation) were tuned to determine the architecture that achieved the lowest test error. The proposed model improved the prediction performance (measured by the root mean square error) by 11-53% for particulate matter, 20-31% for ozone, 9-47% for nitrogenoxides, and 18-46% for sulfurdioxide with respect to the 1-hidden layer long short-term memory networks utilized in the literature. The multivariate model without using meteorological data revealed the best results. Regarding transfer learning, the network weights were transferred from the source city to the target city. The model has more accurate prediction performance with the transfer of the network from Kocaeli to Istanbul as those neighbor cities have similar air pollution and meteorological characteristics.
Bayesian Learning for Cellular Neural Networks
(Kadir Has Üniversitesi, 2013) Ozer, Metin; Ozmen, Atilla
Cellular Neural Networks have been an active research eld since their introduction in the late 80s. Several training algorithms are proposed since then. All have their advantages and disadvantages. Most of them uses deterministic methods to acquire the network parameters. in this thesis a new training method is proposed for Cellular Neural Networks and Discrete-Time Cellular Neural Networks are used for implemented applications. This new method is a probabilistic method. Maximum A Posteriori estimation is used to estimate the network parameters thus making this method a Bayesian learning method. A Cellular Neural Network is nonlinear in the sense of its activation function. For the same reason modeling of a Cellular Neural Network is also nonlinear. Using Maximum A Posteriori estimation on a nonlinear system causes some problems. To cope with this problems in the estimation process of network parameters Metropolis-Hastings algorithm which is one of Monte Carlo Markov Chain methods is used for generating the samples needed from the resulting distribution. After the network is trained it is tested against known algorithms to verify the training process. Discrete-Time Cellular Neural Networks are mostly used for image processing applications. Many dierent kind of applications can be applied using dierent network parameters without changing the cellular network architecture. A couple of applications are picked from this pool and using the estimated parameters Cellular Neural Networks are used to perform some image processing algorithms. This operations are performed by computer models and simulations. -- Abstract'tan.
Citation - WoS: 3
Citation - Scopus: 3
Deep Learning Based Combining Rule for the Estimation of Vapor-Liquid Equilibrium
(Springer Heidelberg, 2023) Bekri, Sezin; Ozmen, Dilek; Ozmen, Atilla
Vapor-liquid equilibrium (VLE) data plays a vital role in the design, modeling and control of process equipment. In this study, to estimate the VLE data of binary systems, a deep neural network (DNN)-based combining rule was proposed based on the cross-term parameter (a(ij)) in the two-parameter Peng-Robinson cubic equation of state (PR-EoS) combined with the one-parameter classical van der Waals mixing and combining rule (1PVDW). Experimental VLE data of alternative binary refrigerant systems selected from the literature were calculated using both the PR + 1PVDW and the DNN-based model. Vapor phase mole fractions (y(i)) and equilibrium pressures (P) obtained from the proposed DNN-based and PR + 1PVDW models were compared in the terms of average percent deviations. For the DNN-based model, the vapor phase mole fractions give at least as good results as the models in the literature, and also it has been shown that a much better estimate of the equilibrium pressure (P) is obtained when compared with that of the literature. Results obtained using the proposed DNN-based model are presented with tables and graphs. For the equilibrium pressure, while the average percent deviation errors (Delta P/P%) calculated in the literature are less than 7.739, the errors obtained with the proposed DNN-based model are smaller than 3.455. And also, for vapor phase mole fractions, while the maximum error (Delta(y1)/(y1) %) in the literature is obtained as 6.142, the largest error calculated with DNN-based model is 3.545. It has been seen that the proposed DNN-based model makes more practical and less error-prone estimations than the methods in the literature.
Performance of Cellular Neural Network Based Channel Equalizers
(2020) Ozmen, Atilla; Enol, H. S; Tander, B.
Abstract—In this paper, a popular dynamic neural network structure called Cellular Neural Network (CNN) is employed as a channel equalizer in digital communications. It is shown that, this nonlinear system is capable of suppressing the effect of intersymbol interference (ISI) and the noise at the channel. The architecture is a small-scaled, simple neural network containing only 25 neurons (cells) with a neighborhood of r = 2 , thus including only 51 weight coefficients. Furthermore, a special technique called repetitive codes in equalization process is also applied to the mentioned CNN based system to show that the two-dimensional structure of CNN is capable of processing such signals, where performance improvement is observed. Simula-tions are carried out to compare the proposed structures with minimum mean square error (MMSE) and multilayer perceptron (MLP) based equalizers.