Pakize Erdogmus

Binary apple tree: A game approach to tree traversal algorithms

The computer science students mostly face with the difficulties in learning the topics of algorithms courses. Only listening the topic from the teacher or just writing makes the learning volatile. Instead of listening or writing, if there is something visual, it would be more permanent to learn because visuality increases the learning potential and the time for learning is minimized. The adversities of classical education techniques were intended to be eliminated in this study via computer games which are becoming more and more popular in this age. An educational convenience is provided for the subject of tree traversal algorithms. Tree traversal algorithms are one of the basic and confused concepts in algorithms and programming courses in computer science. A game called “binary apple tree” was established to teach and learn the subject easier.

Histogram-based automatic segmentation of images

The segmentation process is defined by separating the objects as clustering in the images. The most used method in the segmentation is k-means clustering algorithm. k-means clustering algorithm needs the number of clusters, the initial central points of clusters as well as the image information. However, there is no preliminary information about the number of clusters in real-life problems. The parameters defined by the user in the segmentation algorithms affect the results of segmentation process. In this study, a general approach performing segmentation without requiring any parameters has been developed. The optimum cluster number has been obtained searching the histogram both vertically and horizontally and recording the local and global maximum values. The quite nearly values have been omitted, since the near local peaks are nearly the same objects. Segmentation processes have been performed with k-means clustering giving the possible centroids of the clusters and the optimum cluster number obtained from the histogram. Finally, thanks to histogram method, the number of clusters of k-means clustering has been automatically found for each image dataset. And also, the histogram-based finding of the number of clusters in datasets could be used prior to clustering algorithm for other signal or image-based datasets. These results have shown that the proposed hybrid method based on histogram and k-means clustering method has obtained very promising results in the image segmentation problems.

GPU accelerated training of image convolution filter weights using genetic algorithms

This paper proposes a fast algorithm for training image filter using GPU.Parallelization of genetic algorithms is realized by master-slave method.Sub-image based (SBM) method is proposed to use the GPU efficiently.SBM is developed by discussing other alternative design considerations.Experimental results show about 50i? to 90i? acceleration using GeForce GTX 660. Genetic algorithms (GA) provide an efficient method for training filters to find proper weights using a fitness function where the input signal is filtered and compared with the desired output. In the case of image processing applications, the high computational cost of the fitness function that is evaluated repeatedly can cause training time to be relatively long. In this study, a new algorithm, called sub-image blocks based on graphical processing units (GPU), is developed to accelerate the training of mask weights using GA. The method is developed by discussing other alternative design considerations, including direct method (DM), population-based method (PBM), block-based method (BBM), and sub-images-based method (SBM). A comparative performance evaluation of the introduced methods is presented using sequential and other GPUs. Among the discussed designs, SBM provides the best performance by taking advantage of the block shared and thread local memories in GPU. According to execution duration and comparative acceleration graphs, SBM provides approximately 55-90 times more acceleration using GeForce GTX 660 over sequential implementation on a 3.5GHz processor.

A hybrid dermoscopy images segmentation approach based on neutrosophic clustering and histogram estimation

Abstract In this work, a novel skin lesion detection approach, called HBCENCM, is proposed using histogram-based clustering estimation (HBCE) algorithm to determine the required number of clusters in the neutrosophic c-means clustering (NCM) method. Initially, the dermoscopic images are mapped into the neutrosophic domain over three memberships, namely true, indeterminate, and false subsets. Then, an NCM algorithm is employed to group the pixels in the dermoscopy images, where the number of clusters in the dermoscopy images is determined using the HBCE algorithm. Lastly, the skin lesion is detected based on its intensity and morphological features. The public dataset (ISIC 2016) of 900 images for training and 379 images for testing are used in the present work. A comparative study of the original NCM clustering method is conducted on the same dataset. The results showed the superiority of the proposed approach to detect the lesion with 96.3% average accuracy compared to the average accuracy of 94.6% using the original NCM without HBCE algorithm.

Finding an optimum location for biogas plant: a case study for Duzce, Turkey

This study is a case study for modelling and solving a real-life problem. In this study, a practical approximation for finding an optimum location of a foundation was realized with k-means clustering and optimization. Duzce, in the northwest of Turkey, has been researched for the biogas potential to found biogas plant. With this aim, the number of poultry in Duzce has been determined and presented their potential of biogas. Since the number of poultry is quite enough to found a biogas plant, later the location of the poultry farms and their potentials has been determined. Since there are more than 400 poultry farms in Duzce, firstly locations are clustered with classical k-means algorithm. k is specified as 6–8 with an expert knowledge. Later, the nearest location for each cluster center has been attained with simulated annealing with the objective of minimizing the transportation cost. As a result, it has been determined an optimum location for probable biogas plant for Duzce.

A hybrid approach to image segmentation: Combination of BBO (Biogeography based optimization) and Histogram Based Cluster Estimation

Image segmentation is the process of separating objects within an image. Image segmentation can be considered as an important computer vision problem in image sensing where the homogeneous regions in an image can be distinguished with high accuracy. In this study, a two stage hybrid method has been proposed for image segmentation. In the first stage, the Histogram Based Cluster Estimation (HBCE) is used to determine the number of clusters on the image. In the second stage, the cluster numbers determined by the HBCE algorithm are given to the BBO (Biogeography based optimization) algorithm and then image segmentation is performed. In this study, the proposed hybrid image segmentation method was applied to 6 different images taken from Berkeley database and compared with human segmentation results obtained from the same database. To test the performance of the proposed image segmentation method, RI (Rand Index), GCE (Global Consistency Error) and run time as comparison criterion have been used. The proposed method has been compared with other hybrid methods namely HBCE-PSO (Particle Swarm Optimization) and HBCE-k means clustering. When running on 6 different images, the best Rand Index values from the results obtained for all three methods are as follows; HBCE-BBO incorporation: 0.9859, HBCE-PSO incorporation: 0.9856, HBCE-k means incorporation: 0.7570. The results have shown that the HBCE-BBO hybrid method yields better results than the other two hybrid methods used in working with 6 different image segmentations.

Comparative study of heart disease classification

The aim of this paper is to compare two important machine learning platform results for the same dataset. With this aim, we conducted an experiment to classify heart disease both in Matlab© environment and WEKA©, by using six different algorithms. Linear SVM, Quadratic SVM, Cubic SVM, Medium Gaussian SVM, Decision Tree and Ensemble Subspace Discriminant machine learning approaches are used for classifying the heart disease.

A new approach in human retina optic disc segmentation using Graph Cut

Eyes, which is one of the most significant sense organs, could loose its functionality depending on the environment and growing age. In this study, a method for Optic Dist detection that is one of the first steps in diognising eye diseases and that plays and important role in early diagnosis of many eye diseases have been developed. In this study the red channel is used to eliminate blood vessels in the retina images obtained from the MESSIDOR databese. After this process Graph Cut algorithm is implemented on the retinal imges so the optic disc showing the most significant region is obtained. In this process the morphological operations are applied to improve the image, to eliminate of roughness and to eliminate noise. After this process on image optical disk boundaries is obtained. Recently, this algorithm gained importance due to the results in specially vessel segmentation. In this study, Graph Cut algorithm is used for the first time in optic disc segmentation. In terms of the originality of the algorithm used in this study and the success rate achived, the results are encouraging.

g-BSAFCM: A new hybrid clustering algorithm

Clustering is dividing a dataset into subsets that has similar characteristics. In this study, fuzzy c-means clustering algorithm (FCM) and a new evolutionary optimization algorithm, Backtracking Search (BSA) algorithm, were combined and a new hybrid clustering algorithm (BSAFCM) was proposed. Moreover, the local search abilities of the new algorithm was improved and the new algorithm was named as g-BSAFCM. Three benchmark datasets from UCI Machine Learning Repository database were clustered by using the developed algorithms and FCM. According to the results g-BSAFCM has achieved better results than FCM and BSAFCM.