Gokhan Kayhan

Reviewing and designing pre-processing units for RBF networks: initial structure identification and coarse-tuning of free parameters

This paper reviews some frequently used methods to initialize an radial basis function (RBF) network and presents systematic design procedures for pre-processing unit(s) to initialize RBF network from available input–output data sets. The pre-processing units are computationally hybrid two-step training algorithms that can be named as (1) construction of initial structure and (2) coarse-tuning of free parameters. The first step, the number, and the locations of the initial centers of RBF network can be determined. Thus, an orthogonal least squares algorithm and a modified counter propagation network can be employed for this purpose. In the second step, a coarse-tuning of free parameters is achieved by using clustering procedures. Thus, the Gustafson–Kessel and the fuzzy C-means clustering methods are evaluated for the coarse-tuning. The first two-step behaves like a pre-processing unit for the last stage (or fine-tuning stage—a gradient descent algorithm). The initialization ability of the proposed four pre-processing units (modular combination of the existing methods) is compared with three non-linear benchmarks in terms of root mean square errors. Finally, the proposed hybrid pre-processing units may initialize a fairly accurate, IF–THEN-wise readable initial model automatically and efficiently with a minimum user inference.

The Estimation of Adult and Nymph Stages of Aphis fabae (Hemiptera: Aphididae) Using Artificial Neural Network

ABSTRACT In this research, the estimation of adult and nymph stages and adult of Aphis fabae was investigated using artificial neural network. Determining A. fabae nymph stages is difficult. Morphometric study of different parts of an insects body is needed to obtain an index to distinguish between different immature stages. The study was aimed to develop a model of A. fabae nymph stages and adult using length of hind tibia, antenna and body length. It was found that the constructed artificial neural network (ANN) exhibited high performance for predicting A. fabae nymph stages. Correlation was 99% and the estimation of the best ANN model was determined to be 0.016289 at epoch 18. Software computing techniques are very useful tools for precision agriculture and also determining which method gives the most accurate result.

Performance evaluation of ANN based channel interpolation for OFDM system

In this study, the effect of three different techniques that used for interpolation on OFDM system with pilot based comb type channel estimation is investigated and results are given in terms of BER (Bit Error Rate). The estimation of channel at pilot subcarriers is done by LS (Least Square), and interpolation of channel at data subcarriers are obtained by low pass (LP) interpolation algorithm, ANFIS (Adaptive Network Based Fuzzy Inference Systems) and GRNN (Generalized Regression Neural Networks) artificial neural network (ANN) structures. The results show that there is a relationship between the number of used pilot bits and interpolation techniques performance. If the aim is to get high bandwidth efficiency over a given bandwidth LP algorithm should be used for interpolation, to get the lowest BER ANFIS or GRNN can be used.

DETERMINA TION OF REFLECTANCE VALUES OF HYPERICUM'S LEAVES UNDER STRESS CONDITIONS USING ADAPTIVE NETWORK BASED FUZZY INFERENCE SYSTEM

The effects of water stress and salt levels on hypericums leaves were examined on greenhouse-grown plants of Hypericum perforatum L. by spectral reflectance. Salt levels and irrigation levels were applied 0, 1, 2.5 and 4 deci Siemens per meter (dS/m), 80%, 100% and 120% respectively. Adaptive Network based Fuzzy Inference System (ANFIS) was performed to estimate the effects of water stress and salt levels on spectral reflectance. As a result of ANFIS, it was found that there was close relationship between actual and predicted reflectance values in Hypericum perforatum L. leaves. Performance of ANFIS was examined under different numbers of epoch and rules. On the other hand, RMSE, correlation and analysis time values were found as outputs. Correlation was 99%. The estimation of optimal ANFIS model was determined in 3*3*3 number of rules with 400 epochs.

Using Artificial Neural Network and Multiple Linear Regression for Predicting the Chlorophyll Concentration Index of Saint John’s Wort Leaves

ABSTRACT This research investigates and compares artificial neural network and multiple linear regression for predicting the chlorophyll concentration index of Saint John’s wort leaves (Hypericum perforatum L.). Plants were fertilized with 0, 30, 60, 90, and 120 kg ha−1 nitrogen [34% nitrogen ammonium nitrate (NH4NO3)]. Chlorophyll concentration index of each leaf was measured using SPAD meter. Afterwards, rgb (red, green, and blue color) values of all leaf images were determined by image processing. Values obtained were modeled using both multiple regression analysis and artificial neural networks. Using multiple regression analysis R2 values were between 0.61 and 0.97. Coefficient of determination values (R2) using artificial neutral network values were found to be 0.99. Artificial neutral network modeling successfully described the relationship between actual chlorophyll concentration index values and predicted chlorophyll concentration index values.

Comparision of some models for estimation of reflectance of hypericum leaves under stress conditions

Lack of water resources and high water salinity levels are among the most important growth-restricting factors for plants species of the world. This research investigates the effect of irrigation levels and salinity on reflectance of Saint John’s wort leaves (Hypericum perforatum L.) under stress conditions (water and salt stress) by multiple linear regression (MLR), artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). Empirical and heuristics modeling methods were employed in this study to relate stress conditions to leaf reflectance. It was found that the constructed ANN model exhibited a high performance than multiple regression and ANFIS in estimating leaf reflectance accurately.

Estimation of Chlorophyll Concentration Index at Leaves using Artificial Neural Networks

In this study, the effectiveness of an SPAD-502 portable chlorophyll (Chl) meter was evaluated for estimating the Chl contents in leaves of some medicinal and aromatic plants. To predict the individual chlorophyll concentration indexes of St. John’s wort (Hypericum perforatum L.), mint (Mentha angustifolia L.), melissa (Melissa officinalis L.), thyme (Thymus sp.), and echinacea (Echinacea purpurea L.), models were developed using SPAD value. Multi-layer perceptron (MLP), adaptive neuro fuzzy inference system (ANFIS), and general regression neural network (GRNN) were used for determining the chlorophyll concentration indexes.

Determination of the sources of electric field strength in a medium by artificial neural network

This study aims to determine total electric field strength (E) levels and its main sources in Samsun city center. To reach this goal extensive E and band selective E measurements were performed using Narda SRM 3006. The measurements were collected on the busiest streets of Samsun city during a journey of more than two hours on a vehicle moving with a certain velocity. On the basis of over 700 measurements E and E sources within the band were determined. It is seen from the results that the main sources of E were the base stations that operates at GSM900, GSM1800 and UMTS frequency bands. Because of difficulties encountered in band selective measurements (need for specific and expensive meters); a model that helps to obtain the three band selective Es in terms of E was proposed, and its performance was improved by using multilayer perceptron (MLP) artificial neural network system. With the use of this model; E levels of the three main sources can be determined from E in the environment up to 95% accuracy.

Bird's-eye view images taken plant material and counting

In this study, the recognition of agri-food plants out of the images obtained by the UAV and are intended to implement the counting process. Images obtained with the UAV from plants separation from the background; K-Means (K-Means) with the help of visual elements in the classifier was classified as soil and plants. A better image segmentation and noise in the resulting plants were made to eliminate the morphological filtering. The plants on the noise-free image separation nested data for individual numbers of watershed algorithm was applied. To represent the resulting plant was subjected to the binary image acquisition and counting process. Plant identification methods applied and the counting process accuracy and 87.7% sensitivity, 86.6% were found to implement.