In this paper we classified 4 skin lesions (Melanoma,Melanocytic Nevus, Basal Cell Carcinoma, Benign keratosis) from ISIC 2019 dataset which was published by International Skin Imaging Collabration in 2019. We used InceptionV3 convolutional neural network model for classification. We applied two preprocessing methods: High Dimensional Model Representation (HDMR) and Hilbert Transform. In conclusion we obtained 89% accuracy on classification of Basal Cell Carcinoma using Hilbert Transform. Moreover, we obtained 78% accuracy on
classification of Melanoma using Contrast Enhancement High Dimensional Model Representation (HDMR).

In this study, x band microstrip ring patch antenna was designed and its performance was investigated under various types of feeding methods and dimensions. The patch dimensions and dielectric base thickness were kept constant. The simulations were performed by using Ansys HFSS programme. In simulations, FR-4 and Rogers RT-58880 dielectric base materials were preferred. The bandwidth of the antennas changes between 260 to 1160 MHz and the gain of the antennas are between 4.6 and 9.35 dB.

In recent years, electric vehicles have come to the forefront with decreases in fossil fuel sources, fluctuations in fossil fuel prices, and advances in battery technologies and energy management systems. Today, in many developed countries of the world, both the production of electric vehicles and the production and installation of charging stations to meet the energy needs of these vehicles are being carried out. Along with the installation of electric vehicle charging stations, their impact on the electrical installation in which such systems are integrated has also become important. With the integration of electric vehicle charging stations into the existing electrical installations, on the related installations; The need arises to reconsider voltage drop, current carrying capacity of cables, short-circuit breaking capacity of switchgear equipment, selectivity, and similar electrical phenomenon. In this study, the revision needs of the electrical installation that can be realized in case of the integration of electric vehicle charging station to a sample installation are examined together with the simulation results in Simaris. The review focused on the switchgear equipment and wiring revision needs that may arise in an existing installation with an electric vehicle charging station integrated. In this context; voltage drop, current carrying capacity, switchgear change, short-circuit current, and selectivity concepts are emphasized. Finally; together with the simulation results, the necessity of changing electrical equipment in an installation where the electric vehicle charging station is integrated, in order to ensure healthy and safe operating conditions, is revealed.

Solar Power Plant investments are increasing on a daily basis, as part of a strategy to maximize the use of Turkey's domestic and renewable energy resources. One of the factors affecting the efficiency of a photovoltaic (PV) panel is also surface pollution. It is inevitable to increase the surface pollution of PV panels with environmental factors. Many parameters such as the location of PV panels, fixing patterns, the structural/chemical properties of the pollutants affect the frequency of cleaning periods of the PV panels. In this study, a PV panel surface cleaning robot is developed. It performs autonomously cleaning of substances including dirt, dust on PV panel surfaces. Some features of the developed autonomous robot are; (i) switching to autonomous operating mode if it is needed on rainy days, (ii) pure water-saving thanks to the cleaning decision based on the meteorological data, (iii) determining the cleaning operation based on the dust sensor data, (iv) condition monitoring and control via the developed mobile application. Moreover, the cleaning interval on the PV panel efficiency was investigated. Thus, optimum cleaning intervals were determined in terms of both PV panel energy efficiency and pure water-consuming of the designed robot.

In this study, single phase quasi-Z-source inverter (QZSI) circuit was designed and realized which is an improved version of ZSI and offers a unique power conversion concept by eliminating the conceptual and theoretical limitations of the conventional current and voltage source inverters. Simple boost PWM control technique has been employed to the switches using NUCLEO-F411RE development board since this technique doesn’t involve low-frequency ripples on the passive components of the impedance network and has lower distortions at the output. It has been shown that the developed QZSI circuit can work as a buck-boost converter at different shoot-through duty ratios. At the output of the circuit up to 300W, the AC output voltage is obtained with 85% efficiency. It has been observed that simulation and experimental results carried out in the laboratory environment are compatible.

This study focuses on the fractional version of the FitzHugh-Nagumo (FHN) neuron model. Firstly, the stability analysis of the fractional-order FHN neuron model has been performed and the minimum fractional degree, at which the system could exhibit dynamic behavior, has been determined. Then, the responses of the fractional-order FHN neuron model have been obtained using the Grünwald-Letnikov (G-L) fractional derivative method. This method is one of the methods used in the numerical analysis of the systems that are represented by fractional order. Thanks to the hardware solutions of neuron models; the responses of mathematically defined systems can be obtained in the form of real-time signals, the cell membrane properties of the neurons can be described electromechanically, and the parameters that affect the dynamic behavior of neurons can be associated with the characteristics of the electronic components used in hardware solutions. In this study, the circuit implementation of the fractional-order FHN neuron model is emphasized in order to see the usability of fractional-order calculations in systems that are inspired by biology. In this context, the R-C mimetic circuits have been used instead of classical capacitor elements to compensate for the fractional order in the integrator circuits that are designed by using op-amp, resistor and capacitor elements for the hardware solutions of the differential equations. In the first stage of the design of these R-C imitation circuits, a third-order transfer function has been obtained by the Matsuda approximation method. This obtained transfer function has been transformed into FOSTER-I R-C network and it has been used instead of the classical capacitor element in the integrator blocks of the circuit that is designed by us for the circuit implementation solution of the integer-order FHN neuron model. Thus, an alternative approach for circuit solution of the fractional-order FHN neuron model has been introduced and the verification of this structure has been made by the SPICE circuit simulation.

Protein functions are important for understanding the molecular mechanism of living organisms. Protein structures are used when determining the functions of proteins. Protein functions are mostly used to determine the annotations of uncharacterized protein sequences, to understand the cellular mechanisms of living things, to identify functional changes in genes or proteins that cause disease, and to develop new approaches to prevent, treat and diagnose diseases. Protein functions can be determined effectively by experimental methods. However, experimental methods take time and go through many chemical processes, causing these stages to be slow and costly. In addition to these, the annotations of some proteins whose functional structure and sequence are known cannot be specified due to experimental processes. Due to such reasons and disadvantages, computational-based approaches are needed. Artificial intelligence algorithms are generally used for computational-based applications. In order to predict protein functions with artificial intelligence methods, protein sequences must be mapped with certain mapping methods. In this study, prediction of gene ontology-based protein functions was performed using certain protein mapping techniques. The study consists of four different stages; obtaining protein data, mapping protein sequences, classifying protein functions, and determining the performance of protein mapping techniques. At the end of the study, the best accuracy and AUC score in the biological process category was obtained by the PAM250 protein mapping technique and was calculated as 69% and 88%, respectively. In the cellular component category, the best accuracy and AUC value were obtained by FIBHASH protein mapping technique with 64% and 89%, respectively. In the molecular function category, the best result was obtained with FIBHASH with 64% AUC score and 89% accuracy. It has been observed that the combined use of the proposed artificial intelligence method and protein numerical mapping techniques have an effective role in predicting protein functions.

Information security is related with efforts put in to avoid activities such as unauthorized usage, changing or disseminating of information by having access to these information. This should not be only thought as capturing of information but also as avoiding the violation of particulars such as integrity, availability, and confidentiality. Vulnerability that occurs in any one of these three basic elements will be evaluated as violation of information security. In this study, a multi-level access control method was developed. With the model proposed, in addition to the security policies offered by the Bell-LaPadula access control model, a new set of rules was defined and expanded, and a flexible and dynamic access control model was presented. The developed model being proposed in the study has been applied on data cluster which has been obtained from real life. Performance of the proposed model has been compared with the performances of Traditional Access Control models. When the obtained results were compared, it was observed that object access levels were presented more consistently and quickly with the proposed model.

We present a real-time monocular camera pose estimation algorithm for augmented reality applications. Proposed model is a small convolutional neural network that is trained to directly estimate 6 Degree of Freedom (6-DOF) camera pose from an RGB image. Our model is designed to run on real-time devices with low memory and computation power. Our model can estimate the camera pose in less than 1ms while keeping accuracy comparable to the state-of-the art. This was made possible by employing geometrically sound loss functions and algebraic constraints. Furthermore, we introduce a new synthetic dataset for demonstrating the proposed methods capabilities.

Hand-crafted features are efficient methods for image processing, recognition, and computer vision. However, the advancements in data size and image resolution lead to inconvenience in feature extraction. Moreover, they are unstable, method-dependent, and computationally intensive due to high dimensions. Especially, big data on image datasets causes unpredictable long process. It is a definite necessity to adjust the feature extraction algorithms to computer-assisted methods for image processing. Generative representational learning algorithms have been emerging approaches with the advantages of Deep Learning. In this study, I proposed employing Deep Belief Networks (DBN) for breast cancer diagnosis on ROI images. DBN models were iterated on different image sizes to evaluate the impact of dimensionality on ROI images. The proposed DBN model has achieved performance rates of 96.32%, 96.68%, 95.93%, and 96.40% for accuracy, specificity, sensitivity, and precision, respectively. Consequently, the proposed DBN with detailed representational learning is an efficient and robust algorithm for the classification of breast cancer and healthy tissues on mammograms by the advantage of generative architectures.

The active sub-network detection aims to find a group of interconnected genes of disease-related genes in a protein-protein interaction network. In recent years, several algorithms have been developed for this problem. In this study, the analysis of disease specific sub-network identification programs are evaluated using epilepsy data set. Under the same conditions and with the same data set, 9 different programs are run and results of their Greedy algorithm, Genetic algoritm, Simulated Annealing Algorithm, MCC (Maximal Clique Centrality) algorithm, MCODE (Molecular Complex Detection) algorithm and PEWCC (Protein Complex Detection using Weighted Clustering Coefficent) algorithm are shown. The top scoring 5 modules of each program, are compared using fold enrichment analysis and normalized mutual information. Also, the identified subnetworks are functionally enriched using hypergeometric test and hence, disease associated biological pathways are identified. In addition, running times and features of the programs are comparatively evaluated.

With the technological developments, the amount of data stored in the computer environment is increasing very rapidly. Data analysis has become an important research subject for the correct evaluation of these data and to transform them into useful information. Of course, data play an important role in data analysis. However, model performance is highly dependent on the characteristics of the data. For this reason, it is essential to preprocess them before starting any data analysis process. Data preprocessing creates accurate and useful datasets by overcoming erroneous, incomplete, or other unwanted problems. In this study, papers on data preprocessing in the last 5 years have been researched systematically and it has been observed that widely used preprocessing methods are classified under three main branches: data cleaning, data transformation and data reduction. These methods and various algorithms of them are examined, the frequency of use is presented, and comparisons are made in terms of accuracy performance. As the result of the study shows, when data preprocessing methods are not used on raw data or when wrong data preprocessing methods are applied, data analysis methods alone cannot achieve sufficient performance.

Recent studies show that the performance of Evolutionary Algorithms often depends on choosing appropriate parameter configurations. Thus, researchers have generally tuned these parameters either looking at the similar research areas in the literature or manually, e.g. Grid Search. However, searching the parameter manually is laborious and time-consuming; therefore, meta-optimization techniques have become commonly used methods to adjust parameters of an algorithm. These techniques can be classified in two widespread manners: off-line, tuning parameters of an algorithm before the algorithm initiates, and on-line, tuning the parameters while it is working. In this paper, Bipolar Matching Tendency (BMT) algorithm has been chosen as the selection method of a Genetic Algorithm (GA). The new obtained algorithm is named GA-BMT and has been used for the first time as an online meta-optimizer. In addition, the paper utilizes two search algorithms (Grid Search, Coarse to Fine Search) with three meta-optimization methods (Standard GA, Particle Swarm Optimization, GA-BMT) to investigate the best parameter settings of the Standard GA for 17 test functions, and offers a comparative work by comparing their results. Furthermore, non-parametric statistical tests, Friedman and Wilcoxon Signed Rank, were performed to demonstrate the significance of the results. Based on the all results that achieved, GA-BMT presents a reasonable achievement.

Watermarking is the process of embedding specific data to prove ownership copyright authentication. It is needed whenever media-files are used without proper permission is granted for authentication accuracy. The current watermarking challenge comes from the ownership proof especially as slight tampering occurs on the audio multimedia file which injure the watermarking causing difficulty in its copyright proof. This paper proposes utilizing counting-based secret sharing strategy to allow validation of ownership correctness watermarking even if some of the multimedia audio-file is interfered, as semi-authentication. The research testing run experimentations showed interesting features although this work is still in its early stage. The authentication verification researched secrecy on the audio’s media remarked data altered as on LSB 3 models (1-LSB, 2-LSB, 3-LSB) testing hiding capacity capability as well as PSNR security. Its promising investigation revealed complete data dependency consequences showing real attractive contribution opportunities to be remarked.

Stress is beneficial when a person is focused, awake and alert. However, exposure to high doses of stress harms a person's health. For this reason, it is important to detect stress and begin relief as soon as possible. In this study, soft keyboard typing behaviors with touchscreen panel, gravity, linear acceleration, and gyroscope data obtained from smartphones were examined. It was observed that there was a correlation between the results obtained and typing behaviors and the stress levels of individuals. In this context, an expanded data set was created. In order to detect stress with higher accuracy, a Mahalanobis distance-based outlier detection approach was applied. Subsequently, a structure was created by combining the ReliefF feature selection method and machine learning techniques to identify efficient features and perform classification. The results obtained by cleaning outlier data showed that the created structures achieved success with high accuracy. In addition, outlier detection and cleaning increased the classification success by 1.77 points.