Browsing by Author "Guzelis, Cuneyt"

Now showing 1 - 12 of 12

Citation - WoS: 4
Citation - Scopus: 5
Analysis of Chaotic Dynamics of Chua's Circuit With Lncosh Nonlinearity
(IEEE, 2013) Kocaoglu, Aykut; Karal, Omer; Guzelis, Cuneyt
Chua's circuit, which demonstrates one of the most complicated nonlinear dynamical behaviors, i.e. chaos, contains a three-segment Piecewise Affine (PWA) resistor as the unique nonlinear element. In this study, the non-smooth nonlinearity of Chua's circuit represented by absolute value is approximated with employing the (smooth) lncosh nonlinearity. In contrast to the other smooth approximation, the 1/lambda lncosh (lambda x) approximation has the property of yielding the absolute value nonlinearity |x| as the limit case when lambda parameter goes to infinity. The bifurcation maps and attractors of introduced Chua's circuit obtained for different lambda parameters are presented in the paper in a comparative way. Computer simulations show that lncosh approximation preserves the chaotic behavior and hence provides the possibility of analyzing the behavior of the Chua's circuit by the methods requiring smoothness.
Citation - WoS: 2
Citation - Scopus: 4
A Comparison of Feature Selection Algorithms for Cancer Classification Through Gene Expression Data: Leukemia Case
(IEEE, 2017) Taşçı, Aslı; İnce, Türker; Guzelis, Cuneyt
In this study, three different feature selection algorithms are compared using Support Vector Machines as classifier for cancer classification through gene expression data. The ability of feature selection algorithms to select an optimal gene subset for a cancer type is evaluated by the classification ability of selected genes. A publicly available micro array dataset is employed for gene expression values. Selected gene subsets were able to classify subtypes of the considered cancer type with high accuracies and showed that these feature selection methods were applicable for bio-marker gene selection.
A Convergent Algorithm for a Cascade Network of Multiplexed Dual Output Discrete Perceptrons for Linearly Nonseparable Classification
(Tubitak Scientific & Technical Research Council Turkey, 2014) Genc, Ibrahim; Guzelis, Cuneyt
In this paper a new discrete perceptron model is introduced. The model forms a cascade structure and it is capable of realizing an arbitrary classification task designed by a constructive learning algorithm. The main idea is to copy a discrete perceptron neuron's output to have a complementary dual output for the neuron, and then to select, by using a multiplexer, the true output, which might be 0 or 1 depending on the given input. Hence, the problem of realization of the desired output is transformed into the realization of the selector signal of the multiplexer. In the next step, the selector signal is taken as the desired output signal for the remaining part of the network. The repeated applications of the procedure render the problem into a linearly separable one and eliminate the necessity of using the selector signal in the last step of the algorithm. The proposed modification to the discrete perceptron brings universality with the expense of getting just a slight modification in hardware implementation.
Citation - WoS: 9
Citation - Scopus: 10
Determining the Bistability Parameter Ranges of Artificially Induced Lac Operon Using the Root Locus Method
(Pergamon-Elsevier Science Ltd, 2015) Avcu, N.; Alyuruk, H.; Demir, G. K.; Pekergin, F.; Cavas, L.; Guzelis, Cuneyt
This paper employs the root locus method to conduct a detailed investigation of the parameter regions that ensure bistability in a well-studied gene regulatory network namely, lac operon of Escherichia coli (E. coli). In contrast to previous works, the parametric bistability conditions observed in this study constitute a complete set of necessary and sufficient conditions. These conditions were derived by applying the root locus method to the polynomial equilibrium equation of the lac operon model to determine the parameter values yielding the multiple real roots necessary for bistability. The lac operon model used was defined as an ordinary differential equation system in a state equation form with a rational right hand side, and it was compatible with the Hill and Michaelis-Menten approaches of enzyme kinetics used to describe biochemical reactions that govern lactose metabolism. The developed root locus method can be used to study the steady-state behavior of any type of convergent biological system model based on mass action kinetics. This method provides a solution to the problem of analyzing gene regulatory networks under parameter uncertainties because the root locus method considers the model parameters as variable, rather than fixed. The obtained bistability ranges for the lac operon model parameters have the potential to elucidate the appearance of bistability for E. coli cells in in vivo experiments, and they could also be used to design robust hysteretic switches in synthetic biology. (C) 2015 Elsevier Ltd. All rights reserved.
Citation - WoS: 18
Citation - Scopus: 21
A Dynamical State Feedback Chaotification Method With Application on Liquid Mixing
(World Scientific Publ Co Pte Ltd, 2013) Sahin, Savas; Guzelis, Cuneyt
This paper introduces chaotic reference model-based dynamical state feedback chaotification method which can be applied to any input-state linearizable (nonlinear) system including linear controllable ones as special cases. In the developed method, any chaotic system of arbitrary dimension can be used as the reference model with no need to transform it into a special form, so providing the advantage of exploiting the vast amount of information on chaotic systems and their implementations available in the literature. To demonstrate the potential effective applications of the method, a permanent magnet dc motor is chaotified by the proposed dynamical state feedback as matching the closed loop dynamics to the well-known Chua's chaotic circuit. Then, an impeller mounted on the chaotified dc motor is used for mixing a corn syrup added acid-base mixture. It is observed in a nonintrusive way that mixing actuated by the chaotified dc motor is more efficient than constant and also periodical motor speed cases for the consideration of neutralization time and power consumption together.
Citation - WoS: 7
Citation - Scopus: 7
Exploiting Chaos in Learning System Identification for Nonlinear State Space Models
(Springer, 2015) Olmez, Mehmet; Guzelis, Cuneyt
The paper presents two learning methods for nonlinear system identification. Both methods employ neural network models for representing state and output functions. The first method of learning nonlinear state space is based on using chaotic or noise signals in the training of state neural network so that the state neural network is designed to produce a sequence in a recursive way under the excitement of the system input. The second method of learning nonlinear state space has an observer neural network devoted to estimate the states as a function of the system inputs and the outputs of the output neural network. This observer neural network is trained to produce a state sequence when the output neural network is forced by the same sequence and then the state neural network is trained to produce the estimated states in a recursive way under the excitement of the system input. The developed identification methods are tested on a set of benchmark plants including a non-autonomous chaotic system, i.e. Duffing oscillator. Both proposed methods are observed much superior than well-known identification methods including nonlinear ARX, nonlinear ARMAX, Hammerstein, Wiener, Hammerstein-Wiener, Elman network, state space models with subspace and prediction error methods.
Citation - WoS: 21
Citation - Scopus: 26
Intelligent fashion styling using genetic search and neural classification
(Emerald Group Publishing Ltd, 2015) Vuruşkan, Arzu; İnce, Türker; Bulgun, Ender; Guzelis, Cuneyt
Purpose - The purpose of this paper is to develop an intelligent system for fashion style selection for non-standard female body shapes. Design/methodology/approach - With the goal of creating natural aesthetic relationship between the body shape and the shape of clothing, garments designed for the upper and lower body are combined to fit different female body shapes, which are classified as V, A, H and O-shapes. The proposed intelligent system combines genetic algorithm (GA) with a neural network classifier, which is trained using the particle swarm optimization (PSO). The former, called genetic search, is used to find the optimal design parameters corresponding to a best fit for the desired target, while the task of the latter, called neural classification, is to evaluate fitness (goodness) of each evolved new fashion style. Findings - The experimental results are fashion styling recommendations for the four female body shapes, drawn from 260 possible combinations, based on variations from 15 attributes. These results are considered to be a strong indication of the potential benefits of the application of intelligent systems to fashion styling. Originality/value - The proposed intelligent system combines the effective searching capabilities of two approaches. The first approach uses the GA for identifying best fits to the target shape of the body in the solution space. The second is the PSO for finding optimal (with respect to training mean-squared error) weight and threshold parameters of the neural classifier, which is able to evaluate the fitness of successively evolved fashion styles.
Citation - WoS: 2
Citation - Scopus: 2
Model-Based Robust Chaotification Using Sliding Mode Control
(Tubitak Scientific & Technical Research Council Turkey, 2014) Kocaoglu, Aykut; Guzelis, Cuneyt
Chaos is a complex behavior of dynamical nonlinear systems that is undesirable in most applications and should be controlled; however, it is desirable in some situations and should be generated. In this paper, a robust chaotification scheme based on sliding mode control is proposed for model based chaotification. A continuous time single input observable system is considered such that it is subject to parameter uncertainties, nonlinearities, noises, and disturbances, which are all additive to the input and can be modeled as an unknown function but bounded by a known function. The designed dynamical state feedback control law forces the system to match a reference chaotic system in finite time irrespective of the mentioned uncertainties, noises, and disturbances, as provided by the developed sliding mode control scheme. Simulation results are provided to illustrate the robustness of the proposed scheme against parameter uncertainties and noises. The results are compared with those of other model-based methods and Lyapunov exponents are calculated to show whether the closed-loop control systems exhibit chaotic behavior or not.
Citation - WoS: 89
Citation - Scopus: 123
A New Facial Expression Recognition Based on Curvelet Transform and Online Sequential Extreme Learning Machine Initialized With Spherical Clustering
(Springer London Ltd, 2016) Ucar, Aysegul; Demir, Yakup; Guzelis, Cuneyt
In this paper, a novel algorithm is proposed for facial expression recognition by integrating curvelet transform and online sequential extreme learning machine (OSELM) with radial basis function (RBF) hidden node having optimal network architecture. In the proposed algorithm, the curvelet transform is firstly applied to each region of the face image divided into local regions instead of whole face image to reduce the curvelet coefficients too huge to classify. Feature set is then generated by calculating the entropy, the standard deviation and the mean of curvelet coefficients of each region. Finally, spherical clustering (SC) method is employed to the feature set to automatically determine the optimal hidden node number and RBF hidden node parameters of OSELM by aim of increasing classification accuracy and reducing the required time to select the hidden node number. So, the learning machine is called as OSELM-SC. It is constructed two groups of experiments: The aim of the first one is to evaluate the classification performance of OSELM-SC on the benchmark datasets, i.e., image segment, satellite image and DNA. The second one is to test the performance of the proposed facial expression recognition algorithm on the Japanese Female Facial Expression database and the Cohn-Kanade database. The obtained experimental results are compared against the state-of-the-art methods. The results demonstrate that the proposed algorithm can produce effective facial expression features and exhibit good recognition accuracy and robustness.
Citation - WoS: 1
Citation - Scopus: 1
Numerically Efficient Analysis of a One-Dimensional Stochastic Lac Operon Model
(Springer International Publishing Ag, 2016) Avcu, Neslihan; Pekergin, Nihal; Pekergin, Ferhan; Guzelis, Cuneyt
Gene expression models and their analysis play a key role to understand gene regulation mechanisms. The lac operon mechanism has been largely studied to analyze its bistable behavior. In this paper a stochastic quasi steady-state lac operon model which is indeed a one dimensional birth-death process is considered. Nevertheless the well known closed-from solutions, due to the nonlinearity of parameters, the intermediate computed values become out of the representation range with the increase of the state space size. An aggregation-based two step algorithm is proposed to compute the steady-state distribution efficiently. The results of the stochastic model give the same parameter range for the bistable behavior as with the deterministic ODE model.
Citation - WoS: 9
Citation - Scopus: 14
A Penalty Function Method for Designing Efficient Robust Classifiers With Input Space Optimal Separating Surfaces
(Scientific Technical Research Council Turkey-Tubitak, 2014) Ucar, Aysegsul; Demir, Yakup; Guzelis, Cuneyt
This paper considers robust classification as a constrained optimization problem. Where the constraints are nonlinear, inequalities defining separating surfaces, whose half spaces include or exclude the data depending on their classes and the cost, are used for attaining robustness and providing the minimum volume regions specified by the half spaces of the surfaces. The constraints are added to the cost using penalty functions to get an unconstrained problem for which the gradient descent method can be used. The separating surfaces, which are aimed to be found in this way, are optimal in the input data space in contrast to the conventional support vector machine (SVM) classifiers designed by the Lagrange multiplier method, which are optimal in the (transformed) feature space. Two types of surfaces, namely hyperellipsoidal and Gaussian-based surfaces created by radial basis functions (RBFs), are focused on in this paper due to their generality. Ellipsoidal classifiers are trained in 2 stages: a spherical surface is found in the first stage, and then the centers and the radii found in the first stage are taken as the initial input for the second stage to find the center and covariance matrix parameters of the ellipsoids. The penalty function approach to the design of robust classifiers enables the handling of multiclass classification. Compared to SVMs, multiple-kernel SVMs, and RBF classifiers, the proposed classifiers are found to be more efficient in terms of the required training time, parameter setting time, testing time, memory usage, and generalization error, especially for medium to large datasets. RBF-based input space optimal classifiers are also introduced for problems that are far from ellipsoidal, e.g., 2 Spirals.
Citation - WoS: 19
Citation - Scopus: 21
Segmentation of Abdominal Organs From Mr Images Using Multi-Level Hierarchical Classification
(Gazi Univ, Fac Engineering Architecture, 2015) Selvi, Esref; Selver, M. Alper; Kavur, Ali Emre; Guzelis, Cuneyt; Dicle, Oguz
Medical imaging modalities can provide very detailed and informative mappings of the anatomy of a subject. These detailed and informative mappings can be processed to extract the information of interest instead of dealing with whole data (segmentation). Since manual segmentation on each slice is time consuming, tedious and operator dependent, automatic tools and techniques are needed. Segmentation of abdominal organs is a very challenging field of application due to overlapping intensity ranges of the organs, variations in human anatomy and pathology and the number of studies is very limited for Magnetic Resonance (MR), which is a relatively newer and rapidly developing imaging modality. Since it is obligatory to analyze and visualize MR images of abdominal organs (i.e. liver, right/left kidneys, spleen, pancreas, gall bladder) for several medical procedures, the main goal of this paper is to design and develop a segmentation system (method+software), which is robust to the challenges mentioned above, adaptive to the properties of the abdominal organs as well as to the interrelationships of these organs.