Scopus İndeksli Yayınlar Koleksiyonu / Scopus Indexed Publications Collection
Permanent URI for this collectionhttps://hdl.handle.net/11147/7148
Browse
19 results
Filters
Settings
Search Results
Article Enhancing Thickness Determination of Nanoscale Dielectric Films in Phase Diffraction-Based Optical Characterization Systems With Radial Basis Function Neural Networks(IOP Publishing, 2023) Ataç, Enes; Karatay, Anıl; Dinleyici, Mehmet SalihAccurate determination of the optical properties of ultra-thin dielectric films is an essential and challenging task in optical fiber sensor systems. However, nanoscale thickness identification of these films may be laborious due to insufficient and protracted classical curve matching algorithms. Therefore, this experimental study presents an application of a radial basis function neural network in phase diffraction-based optical characterization systems to determine the thickness of nanoscale polymer films. The non-stationary measurement data with environmental and detector noise were subjected to a detailed analysis. The outcomes of this investigation are benchmarked against the linear discriminant analysis method and further verified by means of scanning electron microscopy. The results show that the neural network has reached a remarkable accuracy of 98% and 82.5%, respectively, in tests with simulation and experimental data. In this way, rapid and precise thickness estimation may be realized within the tolerance range of 25 nm, offering a significant improvement over conventional measurement techniques.Article Citation - WoS: 8Citation - Scopus: 14An End-To Trainable Feature Selection-Forecasting Architecture Targeted at the Internet of Things(IEEE, 2021) Nakıp, Mert; Karakayalı, Kubilay; Güzeliş, Cüneyt; Rodoplu, VolkanWe develop a novel end-to-end trainable feature selection-forecasting (FSF) architecture for predictive networks targeted at the Internet of Things (IoT). In contrast with the existing filter-based, wrapper-based and embedded feature selection methods, our architecture enables the automatic selection of features dynamically based on feature importance score calculation and gamma-gated feature selection units that are trained jointly and end-to-end with the forecaster. We compare the performance of our FSF architecture on the problem of forecasting IoT device traffic against the following existing (feature selection, forecasting) technique pairs: Autocorrelation Function (ACF), Analysis of Variance (ANOVA), Recurrent Feature Elimination (RFE) and Ridge Regression methods for feature selection, and Linear Regression, Multi-Layer Perceptron (MLP), Long Short Term Memory (LSTM), 1 Dimensional Convolutional Neural Network (1D CNN), Autoregressive Integrated Moving Average (ARIMA), and Logistic Regression for forecasting. We show that our FSF architecture achieves either the best or close to the best performance among all of the competing techniques by virtue of its dynamic, automatic feature selection capability. In addition, we demonstrate that both the training time and the execution time of FSF are reasonable for IoT applications. This work represents a milestone for the development of predictive networks for IoT in smart cities of the near future.Article Citation - WoS: 9Citation - Scopus: 11Neural Network-Based Repetitive Learning Control of Euler Lagrange Systems: an Output Feedback Approach(IEEE, 2018) Tatlıcıoğlu, Enver; Çobanoğlu, Necati; Zergeroǧlu, ErkanIn this letter, position tracking control problem of a class of fully actuated Euler Lagrange (EL) systems is aimed. The reference position vector is considered to be periodic with a known period. Only position measurements are available for control design while velocity measurements are not. Furthermore, the dynamic model of the EL systems has parametric and/or unstructured uncertainties which avoid it to be used as part of the control design. To address these constraints, an output feedback neural network-based repetitive learning control strategy is preferred. Via the design of a dynamic model independent velocity observer, the lack of velocity measurements is addressed. To compensate for the lack of dynamic model knowledge, universal approximation property of neural networks is utilized where an online adaptive update rule is designed for the weight matrix. The functional reconstruction error is dealt with the design of a novel repetitive learning feedforward term. The outcome is a dynamic model independent output feedback neural network-based controller with a repetitive learning feedforward component. The stability of the closed-loop system is investigated via rigorous mathematical tools with which semi-global asymptotic stability is ensured. © 2017 IEEE.Article Citation - WoS: 6Citation - Scopus: 7Neural Network Based Robust Control of an Aircraft(ACTA Press, 2020) Tanyer, İlker; Tatlıcıoğlu, Enver; Zergeroǧlu, ErkanOutput tracking control of an aircraft subject to uncertainties in the dynamic model and additive state-dependent nonlinear disturbancelike terms is aimed. Uncertainties in the aircraft dynamic model yield an uncertain input gain matrix, which is neither positive definite nor symmetric and an uncertain term in the error dynamics. To deal with the uncertain input gain matrix, a decomposition method is utilized to put error dynamics in a form where an uncertain positive definite matrix multiplies the auxiliary error but this results in the control input to be pre-multiplied first with a unity upper triangular matrix which is uncertain and then with a known diagonal matrix. A novel controller composed of a neural network compensation term and an integral of signum of error is designed. A novel Lyapunov type stability analysis is utilized to prove global asymptotic tracking of output of a reference model. Extensive numerical simulations are presented to demonstrate the efficacy of the proposed controller where robustness to variation of initial states and a comparison with a robust controller are also shown. © 2020 Acta Press. All rights reserved.Article Citation - WoS: 4Citation - Scopus: 8Fault Diagnosis of a Wind Turbine Simulated Model Via Neural Networks(IFAC Secretariat, 2018) Simani, Silvio; Turhan, CihanThe fault diagnosis of wind turbine systems has been proven to be a challenging task and motivates the research activities carried out through this work. Therefore, this paper deals with the fault diagnosis of wind turbines, and it proposes viable solutions to the problem of earlier fault detection and isolation. The design of the fault indicator involves a data-driven approach, as it represents an effective tool for coping with a poor analytical knowledge of the system dynamics, together with noise and disturbances. In particular, the data-driven proposed solution relies on neural networks that are used to describe the strongly nonlinear relationships between measurement and faults. The chosen network architecture belongs to the nonlinear autoregressive with exogenous input topology, as it can represent a dynamic evolution of the system along time. The developed fault diagnosis scheme is tested by means of a high-fidelity benchmark model, that simulates the normal and the faulty behaviour of a wind turbine. The achieved performances are compared with those of other control strategies, coming from the related literature. Moreover, a Monte Carlo analysis validates the robustness of the proposed solutions against the typical parameter uncertainties and disturbances.Conference Object Citation - Scopus: 6Gender Prediction From Tweets With Convolutional Neural Networks: Notebook for Pan at Clef 2018(CEUR Workshop Proceedings, 2018) Sezerer, Erhan; Polatbilek, Ozan; Sevgili, Özge; Tekir, SelmaThis paper presents a system1 developed for the author profiling task of PAN at CLEF 2018. The system utilizes style-based features to predict the gender information from the given tweets of each user. These features are automatically extracted by Convolutional Neural Networks (CNN). The system mainly depends on the idea that the informativeness of each tweet is not the same in terms of the gender of a user. Thus, the attention mechanism is included to the CNN outputs in order to discriminate the tweets carrying more information. Our architecture was able to obtain competitive results on three languages provided by the PAN 2018 author profiling challenge with an average accuracy of 75.1% on local runs and 70.23% on the submission run.Article Citation - WoS: 1Citation - Scopus: 2Discrimination of Bio-Crystallogram Images Using Neural Networks(Springer Verlag, 2014) Ünlütürk, Sevcan; Ünlütürk, Mehmet S.; Pazır, Fikret; Kuşçu, AlperThis study utilized a unique neural network model for texture image analysis to differentiate the crystallograms from pairs of fresh red pepper fruits from conventional and organic farms. The differences in visually analyzed samples are defined as the distribution of crystals on the circular glass underlay, the thin or thick structure of crystal needles, the angles between branches and side needles, etc. However, the visual description and definition of bio-crystallogram images has major disadvantages. A novel methodology called an image neural network (INN) has been developed to overcome these shortcomings. The 1,488 × 2,240 pixel bio-crystallogram images were acquired in a lab and cropped to 425 × 1,025 pixel images. These depicted either a conventional sweet red pepper or an organic sweet red pepper. A set of 19 images was utilized to train the image neural network. A new set of 4 images was then prepared to test the INN performance. Overall, the INN achieved an average recognition performance of 100 %. This high level of recognition suggests that the INN is a promising method for the discrimination of bio-crystallogram images. In addition, Hinton diagrams were utilized to display the optimality of the INN weights.Article Citation - WoS: 9Citation - Scopus: 10Modelling Trip Distribution With Fuzzy and Genetic Fuzzy Systems(Taylor and Francis Ltd., 2013) Kompil, Mert; Çelik, Hüseyin MuratThis paper explores the potential capabilities of fuzzy and genetic fuzzy system approaches in urban trip distribution modelling with some new features. First, a simple fuzzy rule-based system (FRBS) and a novel genetic fuzzy rule-based system [GFRBS: a fuzzy system improved by a knowledge base learning process with genetic algorithms (GAs)] are designed to model intra-city passenger flows for Istanbul. Subsequently, their accuracy, applicability and generalizability characteristics are evaluated against the well-known gravity- and neural network (NN)-based trip distribution models. The overall results show that: traditional doubly constrained gravity models are still simple and efficient; NNs may not show expected performance when they are forced to satisfy trip constraints; simply-designed FRBSs, learning from observations and expertise, are both efficient and interpretable even if the data are large and noisy; and use of GAs in fuzzy rule-based learning considerably increases modelling performance, although it brings additional computation cost.Article Citation - WoS: 91Citation - Scopus: 105Measurement of the Single-Top T-Channel Cross Section in Pp Collisions at ?s=7 Tev(Springer Verlag, 2012) CMS Collaboration; Karapınar, GülerA measurement of the single-top-quark t-channel production cross section in pp collisions at √s=7 TeV with the CMS detector at the LHC is presented. Two different and complementary approaches have been followed. The first approach exploits the distributions of the pseudorapidity of the recoil jet and reconstructed top-quark mass using background estimates determined from control samples in data. The second approach is based on multivariate analysis techniques that probe the compatibility of the candidate events with the signal. Data have been collected for the muon and electron final states, corresponding to integrated luminosities of 1.17 and 1.56 fb-1, respectively. The single-top-quark production cross section in the t-channel is measured to be 67.2±6.1 pb, in agreement with the approximate next-to-next-to-leading- order standard model prediction. Using the standard model electroweak couplings, the CKM matrix element |V tb| is measured to be 1.020 ± 0.046 (meas.) ± 0.017 (theor.). © 2012 CERN for the benefit of the CMS collaboration.Article Citation - WoS: 12Citation - Scopus: 13The Discrimination of Raw and Uht Milk Samples Contaminated With Penicillin G and Ampicillin Using Image Processing Neural Network and Biocrystallization Methods(Academic Press Inc., 2013) Ünlütürk, Sevcan; Pelvan, Merve; Ünlütürk, Mehmet S.This paper utilized a neural network for texture image analysis to differentiate between milk, either raw or ultra high temperature (UHT) with antibiotic residues (e.g., penicillin G and ampicillin) and milk without antibiotic residues. The biocrystallization method was applied to obtain biocrystallogram images for milk samples spiked with penicillin G and ampicillin at different concentration levels. The biocrystallogram images were used as an input for a designed neural network called the image processing neural network (ImgProcNN). The visual differences in these images that were based on textural properties, including the distribution of crystals on the circular grass underlay, the thin or thick structure of the crystal needles, and the angles between the branches and the side needles, were used to discriminate the antibiotic-free milk samples from samples with antibiotic residues. The visual description and definition of these images have major disadvantages. In this study, the ImgProcNN was developed to overcome the shortcomings of these visual descriptions and definitions. Overall, the neural network achieved an average recognition performance between 86% and 100%. This high level of recognition suggests that the neural network used in this paper has potential as a method for discriminating raw and UHT milk samples contaminated with different antibiotics.