International Journal of Neural Networks and Advanced Applications


ISSN: 2313-0563
Volume 4, 2017

Notice: As of 2014 and for the forthcoming years, the publication frequency/periodicity of NAUN Journals is adapted to the 'continuously updated' model. What this means is that instead of being separated into issues, new papers will be added on a continuous basis, allowing a more regular flow and shorter publication times. The papers will appear in reverse order, therefore the most recent one will be on top.

Main Page

Submit a paper | Submission terms | Paper format

 


Volume 4, 2017


Title of the Paper: Deep Leaning Neural Networks for Determining Replacement Timing of Steel Water Transmission Pipes

 

Authors: Hong Jin Jun, Jae K. Park, Cheol Ho Bae

Pages: 57-63

Abstract: Water main pipe breaks are an ongoing concern worldwide. Large-diameter steel water transmission mains (WTMs) transport a much larger volume of water and their failure leads to even greater damages than those seen in water networks with small diameter iron or PVC pipe lines. However, there is no predictive model for large-diameter steel WTMs, leaving retroactive maintenance as the sole means of prevention. The objective of this study was to predict the optimal replacement timing for large-diameter steel WTMs based on physical and environmental factors, using Deep Learning algorithms. The model was developed in four steps: (1) determine major factors, (2) determine the best model by comparing performances of three neural networks (NNs) (a shallow artificial NN, multiple hidden layered NN, Stacked autoencoder NN), (3) classify the data into homogeneous groups by an ANN-based clustering technique, and (4) perform the developed model for each group. The multiple hidden layered NN was found to be the best deep neural NN in forecasting a replacement timing of aging WTMs. Additionally, it is recommended that such ANN-based clustering methods be used in predicting a more accurate replacement timing of water networks and making a quantitative decision on replacement.


Title of the Paper: Applications of Tactile Sensors and Displays in Robotically-Assisted Minimally Invasive Surgery

 

Authors: Sasha Mbabazi, Javad Dargahi, Amir Molaei

Pages: 51-56

Abstract: Robotically Assisted Minimally Invasive Surgery (RMIS) applications have been shown to help with improving patient recovery times, reducing the duration of operation as well as minimizing trauma in patients. Recent years have seen an increase in the development of devices and external systems with the improve RMIS procedures. There is need for this kind of improvement as the procedure comes with drawbacks to the effectiveness of the surgeon. These could include impaired vision and haptic feedback otherwise available through direct manipulation. Various sensors and tactile display systems possess the potential to solve such problems and thereby further improve RMIS procedure. This paper makes a review of recent and past tactile technology with respect to RMIS applications from early endeavors designed to complement commercially available surgical systems to novel developments yet to be fully evaluated. In addition, a range of suggestions for areas on which to improve and directions for future research are provided.


Title of the Paper: A New Optimization Algorithm for Image Classification Based on the Support Vector Machine

 

Authors: Astrit Hulaj, Adrian Shehu, Xhevahir Bajrami

Pages: 45-50

Abstract: Today, with the aim of improving cross-border security, one of the technologies that today many countries are thinking to apply is WMSN. Applying this technology enables gathering information about potential activities along the border through captured images. This will make it easier for security authorities to detect and identify illegal activities along the green borderline. However, images captured by multimedia sensors can often contain information that aren't with interest for state security authorities. In this paper we will present an algorithm that will enable the classification and identification of images captured by multimedia sensors. The captured images will be sent through the network to the security authorities at the monitoring center, and after the classification are sent to border patrols. The working principle of this algorithm is based on the Suport Vector Machine (SVM).


Title of the Paper: Design and Implementation of a Neural Network for Voiced/Unvoiced Classification for a Given System

 

Authors: Kevin Struwe

Pages: 39-44

Abstract: Voiced/unvoiced classification is a task from the field of acoustics to assess the vocal folds’ contribution to speech production within a given piece of sound. However, it is a difficult task, commonly approached through means of digital signal processing, which usually delivers subpar results, especially in the transition regions between the two classes. Artificial neural networks deliver results of better quality while being able to be more efficient. This paper provides best practices for the design and the implementation of an artificial neural network approach which is able to achieve better results for this particular problem . It outlines the steps to implement a multi-layer perceptron trained with back-propagation using minibatch stochastic gradient descent. The implementation was done in Octave/Matlab.


Title of the Paper: New Dermatological Asymmetry Measure of Skin Lesions

 

Authors: Piotr Milczarski, Zofia Stawska, Łukasz Wąs, Sławomir Wiak, Marek Kot

Pages: 32-38

Abstract: In the paper, we introduce a new dermatological asymmetry measure, DASM. We analyze and present dermatologists’ approaches to melanocytic and non-melanocytic skin lesions. The asymmetry of skin lesion is described as a function of a shape, hue and structure. In the paper, the shape symmetry/asymmetry of the segmented skin lesion is discussed and a new dermatological measure is defined. The results of the method/algorithm using PH2 dataset are presented. Mild and malignant skin lesions are often very difficult to diagnose because there are many factors that can lead to misdiagnosis, which often leads to very long and expensive clinical treatment


Title of the Paper: Side Weir Simulation Using Two Different Support Vector Machine Methods

 

Authors: Hossein Bonakdari, Amir Hossein Zaji

Pages: 28-31

Abstract: Determining the accurate discharge coefficient is a crucial process in side weirs design. Because of the higher performance modified-shape side weirs are used extensively in practical situations. The discharge coefficients of modified-shape side weirs are complex because they are related to various geometric and hydraulic conditions, such as the weir’s height (w), included angle (θ), length (L), and the upstream Froude number (F1). In this study, Support Vector Regression (SVR) was used to predict the discharge coefficient of a modified-shape, labyrinth side weir. Polynomial and radial basis functions were investigated as kernel functions. Instead of minimizing the training error, the Polynomial SVR (Poly-SVR) and the Radial Basis Function SVR (RBF-SVR) minimize the generalization error in the training process. Investigation of the performance of the proposed method showed that both Poly-SVR and RBF-SVR provided accurate predictions, and the generalization-based SVR was used successfully in complex hydraulic prediction problems.


Title of the Paper: Low-Power Stereo Vision Accelerator for Automotive

 

Authors: Mihaela Malita, Octavian Nedescu, Alexandru Negoita, Gheorghe M. Stefan

Pages: 22-27

Abstract: Various forms of Convolutional Neural Network (CNN) architectures are used as Machine Learning (ML) tools for learning the similarity measure on video patches in order to run the stereo matching algorithm – the most computationally intensive stage of the pipeline for the stereo vision function used in designing an autonomous car. We propose a hybrid system for real-time, low-power and high-temperature implementations of the algorithm. The accelerator part of the system is a programmable many-core system with a Map-Reduce Architecture. Our paper describes and evaluates the proposed accelerator for running versions of the stereo matching algorithm.


Title of the Paper: Euclidean & Geodesic Distance Between a Facial Feature Points in Two-Dimensional Face Recognition System

 

Authors: Rachid Ahdid, Khaddouj Taifi, Said Safi, Bouzid Manaut

Pages: 14-21

Abstract: In this paper, we present two feature extraction methods for two-dimensional face recognition. Our approaches are based on facial feature points detection then compute the Euclidean Distance between all pairs of this points for a first method (ED-FFP) and Geodesic Distance in the second approach (GD-FFP). These measures are employed as inputs to a commonly used classification techniques such as Neural Networks (NN), k-Nearest Neighbor (KNN) and Support Vector Machines (SVM). To test the present methods and evaluate its performance, a series of experiments were performed on two-dimensional face image databases (ORL and Yale). The recognition rate across all trials was higher using Geodesic Distance (GD-FFP) than Euclidean Distance (ED-FFP). The experimental results also indicated that the extraction of image features is computationally more efficient using Geodesic Distance than Euclidean Distance.


Title of the Paper: Solving Problems for New Results Predictions in Artificial Neural Networks

 

Authors: Imanol Bilbao, Javier Bilbao

Pages: 10-13

Abstract: Within Machine Learning field, Artificial Neural Networks (ANN) have taken a new boost for researching and applications. ANNs are configured (or learn) to solve a certain problem. The term supervised learning refers to algorithms which find a mapping between a set of inputs called features and the provided output values. Optimization problems are one of the fields where ANN have been developed successfully. But one of the problems that a developer must solve when he design a new neural network is the called overfitting. In this paper, we study this problem and how to solve it.


Title of the Paper: Numerical Simulation for ANN Training and Validation for Impact Detection

 

Authors: M. Viscardi, P. Napolitano

Pages: 1-9

Abstract: In last decade the presence of composite structures is increased dramatically. Today the composite is present in every field of our life. A lot of objects have a percentage of composites material inside. With the last release of Boeing 787 Dreamliner the composite is become relevant also in aerospace field. Composite materials have a lot of benefit, first of all they can merge high strength and low weight. The life cycle of composite material is far away from metal’s one. They need a better monitoring to avoid the delamination problem that is strictly related to objects impact. To help the monitoring and increase the safety of composite structure an artificial neural network has been developed with the task to catch impacts on composite structures and give informations about the integrity status.