Okan Ersoy - Academia.edu (original) (raw)

Papers by Okan Ersoy

Sigma Journal of Engineering and Natural Sciences, Mar 1, 2018

Cornell University - arXiv, Nov 28, 2012

Graph embedding techniques are useful to characterize spectral signature relations for hyperspect... more Graph embedding techniques are useful to characterize spectral signature relations for hyperspectral images. However, such images consists of disjoint classes due to spatial details that are often ignored by existing graph computing tools. Robust parameter estimation is a challenge for kernel functions that compute such graphs. Finding a corresponding high quality coordinate system to map signature relations remains an open research question. We answer positively on these challenges by first proposing a kernel function of spatial and spectral information in computing neighborhood graphs. Secondly, the study exploits the force field interpretation from mechanics and devise a unifying nonlinear graph embedding framework. The generalized framework leads to novel unsupervised multidimensional artificial field embedding techniques that rely on the simple additive assumption of pair-dependent attraction and repulsion functions. The formulations capture long range and short range distance ...

Most edge detection algorithms include three main stages: smoothing, differentiation, and labelin... more Most edge detection algorithms include three main stages: smoothing, differentiation, and labeling. In this paper, we evaluate the performance of algorithms in which competitive learning is applied first to enhance edges, followed by an edge detector to locate the edges. In this way, more detailed and relatively more unbroken edges can be found as compared to the results when an edge detector is applied alone. The algorithms compared are K-Means, SOM and SOGR for clustering, and Canny and GED for edge detection. Perceptionally, best results were obtained with the GED-SOGR algorithm. The SOGR is also considerably simpler and faster than the SOM algorithm. 1.

In multi-objective optimization problems, the objective space of fitness functions has a close re... more In multi-objective optimization problems, the objective space of fitness functions has a close relationship with the solution space. Extracting the optimal direction and optimal parameter information are very useful for the optimization process. This paper proposes multi-objective differential evolution algorithm with a clustering based objective space division and parameter adaptation (MODECD). L∞ metric matrix based optimal strategy is used to split the objective space into sub-spaces and to extract the optimal directions. A fitness value based parameter adaptation and mutation strategy are used to extract the optimal strategy information. The results with 20 benchmark tests show the competitiveness of the MODECD algorithm in both convergence speed and diversity of solution approximating the Pareto front. In addition, MODECD is used to optimize the fermentation process of sodium gluconate as an example of its superior performance in solving real-world problems.

Academic Platform Journal of Engineering and Science, 2021

In remote sensing, which is becoming increasingly important today, researchers use high-dimension... more In remote sensing, which is becoming increasingly important today, researchers use high-dimensional data representing the surface of the earth to find relationships between various spectral signatures. In particular, images can consist of hundreds of high-resolution bands that reflect the properties of different materials. However, the presence of a large number of different bands in high-dimensional space can make interpretation of these features difficult. Various difficulties are encountered due to dimensionality problem for pre-processing of remote sensing data. Research in this area reveals that this is a difficult problem and not a single solution to all problems. However, recent studies show that manifold learning techniques are a very important solution in the preprocessing of hyperspectral images. In this study, the performance of the state-of-the-art manifold embedding methods on hyperspectral data is analyzed comparatively. The dimension reduction application of each meth...

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES, 2019

International Journal of Hybrid Information Technology, 2017

John Wiley & Sons, Inc. eBooks, May 18, 2006

$Research paper thumbnail of New techniques for the design of diffractive optical devices for optical communications and networking$

ECE Technical Reports, 2001

Three new techniques are proposed for the design of optical devices for various applications in o... more Three new techniques are proposed for the design of optical devices for various applications in optical communications and networking. The first algorithm shows that the nonperiodic grating-assisted directional coupler design can achieve not only the complete ...

Remote Sensing

Hyperspectral remote sensing image (HSI) classification is very useful in different applications,... more Hyperspectral remote sensing image (HSI) classification is very useful in different applications, and recently, deep learning has been applied for HSI classification successfully. However, the number of training samples is usually limited, causing difficulty in use of very deep learning models. We propose a wide and deep Fourier network to learn features efficiently by using pruned features extracted in the frequency domain. It is composed of multiple wide Fourier layers to extract hierarchical features layer-by-layer efficiently. Each wide Fourier layer includes a large number of Fourier transforms to extract features in the frequency domain from a local spatial area using sliding windows with given strides.These extracted features are pruned to retain important features and reduce computations. The weights in the final fully connected layers are computed using least squares. The transform amplitudes are used for nonlinear processing with pruned features. The proposed method was ev...

ArXiv, 2018

We used the Ladder Network [Rasmus et al. (2015)] to perform Hyperspectral Image Classification i... more We used the Ladder Network [Rasmus et al. (2015)] to perform Hyperspectral Image Classification in a semi-supervised setting. The Ladder Network distinguishes itself from other semi-supervised methods by jointly optimizing a supervised and unsupervised cost. In many settings this has proven to be more successful than other semi-supervised techniques, such as pretraining using unlabeled data. We furthermore show that the convolutional Ladder Network outperforms most of the current techniques used in hyperspectral image classification and achieves new state-of-the-art performance on the Pavia University dataset given only 5 labeled data points per class.

This paper presents a multi-offspring genetic algorithm (MGA) with two-point crossover in accorda... more This paper presents a multi-offspring genetic algorithm (MGA) with two-point crossover in accordance with biology and mathematical ecological theory. For the MGA, the main existing problems are generation methods of multi-offsprings with different crossover methods, the best number of offsprings and the influence of the number of offsprings on the speed of computation. To solve these problems, the paper first studies the relationship between the number of offsprings and the computational speed of the MGA with two-point crossover. Furthermore, the relationship between the generation method of multi-offsprings, the number of offsprings and the computational speed is analyzed. The results with ten test functions show that when the number of offsprings generated by the MGA based on two-point crossover equals 6, the MGA with two-point crossover has significantly improved the computational speed and reduced the number of iterations as compared to the basic genetic algorithm (BGA) and the ...

2005 13th European Signal Processing Conference, 2005

ArXiv, 2020

Parallel implementations of distributed global optimization (DGO) [13] on MP-1 and NCUBE parallel... more Parallel implementations of distributed global optimization (DGO) [13] on MP-1 and NCUBE parallel computers revealed an approximate O(n) increase in the performance of this algorithm. Therefore, the implementation of the DGO on parallel processors can remedy the only draw back of this algorithm which is the O(n) of execution time as the number of the dimensions increase. The speed up factor of the parallel implementations of DGO is measured with respect to the sequential execution time of the identical problem on SPARC IV computer. The best speed up was achieved by the SIMD implementation of the algorithm on the MP-1 with the total speedup of 126 for an optimization problem with n = 9. This optimization problem was distributed across 128 PEs of Mas-Par.

Remote. Sens., 2021

Recently, deep learning methods, for example, convolutional neural networks (CNNs), have achieved... more Recently, deep learning methods, for example, convolutional neural networks (CNNs), have achieved high performance in hyperspectral image (HSI) classification. The limited training samples of HSI images make it hard to use deep learning methods with many layers and a large number of convolutional kernels as in large scale imagery tasks, and CNN-based methods usually need long training time. In this paper, we present a wide sliding window and subsampling network (WSWS Net) for HSI classification. It is based on layers of transform kernels with sliding windows and subsampling (WSWS). It can be extended in the wide direction to learn both spatial and spectral features more efficiently. The learned features are subsampled to reduce computational loads and to reduce memorization. Thus, layers of WSWS can learn higher level spatial and spectral features efficiently, and the proposed network can be trained easily by only computing linear weights with least squares. The experimental results...

Interest in manifold learning for representing the topology of large, high dimensional nonlinear ... more Interest in manifold learning for representing the topology of large, high dimensional nonlinear data sets in lower, but still meaningful dimensions for visualization and classification has grown rapidly over the past decade, and particularly in analysis of hyperspectral imagery. High spectral resolution and the typically continuous bands of hyperspectral image (HSI) data enable discrimination between spectrally similar targets of interest, provide capability to estimate within pixel abundances of constituents, and allow direct exploitation of absorption features in predictive models. The spectral response of the narrow bands is often nonlinear and includes the effects of multipath scattering, localized differences in bidirectional reflectance, and non-uniform attenuation that are often exhibited in remote sensing applications [1]. Because of the dense spectral sampling of HSI data, the associated spectral information in many adjacent bands is highly correlated, resulting in much lo...

Recently, deep learning has been successfully and widely used in hyperspectral image (HSI) classi... more Recently, deep learning has been successfully and widely used in hyperspectral image (HSI) classification. Considering the difficulty of acquiring HSIs, there are usually a small number of pixels used as the training instances. Therefore, it is hard to fully use the advantages of deep learning networks; for example, the very deep layers with a large number of parameters lead to overfitting. This paper proposed a dynamic wide and deep neural network (DWDNN) for HSI classification, which includes multiple efficient wide sliding window and subsampling (EWSWS) networks and can grow dynamically according to the complexity of the problems. The EWSWS network in the DWDNN was designed both in the wide and deep direction with transform kernels as hidden units. These multiple layers of kernels can extract features from the low to high level, and because they are extended in the wide direction, they can learn features more steadily and smoothly. The sliding windows with the stride and subsampl...

Studies on drug design datasets are continuing to grow. These datasets are usually known as hard ... more Studies on drug design datasets are continuing to grow. These datasets are usually known as hard modeled, having a large number of features and a small number of samples. The most common problems in the drug design area are of regression type. Committee machines (ensembles) have become popular in machine learning because of their high performance. In this study, dynamics of ensembles on regression related drug design problems are investigated on a big dataset collection. The study tries to determine the most successful ensemble algorithm, the base algorithm-ensemble pair having the best / worst results, the best successful single algorithm, and the similarities of algorithms according to their performances. We also discuss whether ensembles always generate better results than single algorithms.