scholarly journals Towards Accurate Diagnosis of Skin Lesions Using Feedforward Back Propagation Neural Networks

Diagnostics ◽  
2021 ◽  
Vol 11 (6) ◽  
pp. 936
Author(s):  
Simona Moldovanu ◽  
Cristian-Dragos Obreja ◽  
Keka C. Biswas ◽  
Luminita Moraru
Keyword(s):  
Test Performance ◽  
Structural Design ◽  
Back Propagation ◽  
Skin Lesions ◽  
Public Image ◽  
Data Sets ◽  
Dysplastic Nevus ◽  
Melanoma Detection ◽  
Hidden Layer ◽  
Propagation Network

In the automatic detection framework, there have been many attempts to develop models for real-time melanoma detection. To effectively discriminate benign and malign skin lesions, this work investigates sixty different architectures of the Feedforward Back Propagation Network (FFBPN), based on shape asymmetry for an optimal structural design that includes both the hidden neuron number and the input data selection. The reason for the choice of shape asymmetry was based on the 5–10% disagreement between dermatologists regarding the efficacy of asymmetry in the diagnosis of malignant melanoma. Asymmetry is quantified based on lesion shape (contour), moment of inertia of the lesion shape and histograms. The FFBPN has a high architecture flexibility, which indicates it as a favorable tool to avoid the over-parameterization of the ANN and, equally, to discard those redundant input datasets that usually result in poor test performance. The FFBPN was tested on four public image datasets containing melanoma, dysplastic nevus and nevus images. Experimental results on multiple benchmark data sets demonstrate that asymmetry A2 is a meaningful feature for skin lesion classification, and FFBPN with 16 neurons in the hidden layer can model the data without compromising prediction accuracy.

An intelligent system for selection of grinding wheels

1997 ◽  
Vol 211 (8) ◽  
pp. 635-641 ◽  
Author(s):  
Y Li ◽  
B Mills ◽  
W B Rowe
Keyword(s):  
Neural Network ◽  
Intelligent System ◽  
Back Propagation ◽  
Grinding Wheel ◽  
Neural Network System ◽  
Grinding Conditions ◽  
Training Examples ◽  
Using Data ◽  
Hidden Layer ◽  
Propagation Network

This paper describes the development of a neural network system for grinding wheel selection. The system employs a back-propagation network with one hidden layer and was trained using data from reference handbooks. It is shown that a neural network is capable of learning the relationship between the wheel and the grinding process without a requirement for rules or equations. It was further found that a relatively small number of training examples allows the system to produce reliable recommendations for a much greater number of combinations of grinding conditions. The system was developed on a PC using the C++ programming language.

EFFECTS OF NONSINGULAR PREPROCESSING ON FEEDFORWARD NETWORK TRAINING

2005 ◽  
Vol 19 (02) ◽  
pp. 217-247 ◽  
Author(s):  
CHANGHUA YU ◽  
MICHAEL T. MANRY ◽  
JIANG LI
Keyword(s):  
Back Propagation ◽  
Original Data ◽  
Data Sets ◽  
Training Algorithm ◽  
Feedforward Network ◽  
Data Set ◽  
Network Training ◽  
The Neural Network ◽  
Hidden Layer ◽  
Theoretical Analyses

In the neural network literature, many preprocessing techniques, such as feature de-correlation, input unbiasing and normalization, are suggested to accelerate multilayer perceptron training. In this paper, we show that a network trained with an original data set and one trained with a linear transformation of the original data will go through the same training dynamics, as long as they start from equivalent states. Thus preprocessing techniques may not be helpful and are merely equivalent to using a different weight set to initialize the network. Theoretical analyses of such preprocessing approaches are given for conjugate gradient, back propagation and the Newton method. In addition, an efficient Newton-like training algorithm is proposed for hidden layer training. Experiments on various data sets confirm the theoretical analyses and verify the improvement of the new algorithm.

Comparison of BP and GRNN Algorithm for Factory Monitoring

2011 ◽  
Vol 52-54 ◽  
pp. 2105-2110 ◽  
Author(s):  
Ing Jiunn Su ◽  
Chia Chih Tsai ◽  
Wen Tsai Sung
Keyword(s):  
Neural Network ◽  
Back Propagation ◽  
Predictive Ability ◽  
Network Models ◽  
General Regression Neural Network ◽  
Data Sets ◽  
Artificial Neural ◽  
Propagation Network ◽  
General Regression ◽  
Monitoring Area

Artificial neural networks (ANNs) are one of the most recently explored advanced technologies which show promise in the factory monitoring area. This paper focuses on two particular network models, back-propagation network (BPN) and general regression neural network (GRNN). The prediction accuracy of these two models is evaluated using a practical application situation in a monitor factory. GRNN emerged as a variant of the artificial neural network. Its principal advantages are that it can quickly learn and rapidly converge to the optimal regression surface with large number of data sets. According the simulation results we can show that GRNN is an effective way to considerably improve the predictive ability of BPN.

Comparison of the Hybrid Credit Scoring Models Based on Various Classifiers

2012 ◽  
pp. 220-239
Author(s):  
Fei-Long Chen ◽  
Feng-Chia Li
Keyword(s):  
Nearest Neighbor ◽  
Credit Scoring ◽  
Back Propagation ◽  
Support Vector ◽  
Data Sets ◽  
K Nearest Neighbor ◽  
Hybrid Features ◽  
Credit Data ◽  
Learning Machine ◽  
Propagation Network

Credit scoring is an important topic for businesses and socio-economic establishments collecting huge amounts of data, with the intention of making the wrong decision obsolete. In this paper, the authors propose four approaches that combine four well-known classifiers, such as K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Back-Propagation Network (BPN) and Extreme Learning Machine (ELM). These classifiers are used to find a suitable hybrid classifier combination featuring selection that retains sufficient information for classification purposes. In this regard, different credit scoring combinations are constructed by selecting features with four approaches and classifiers than would otherwise be chosen. Two credit data sets from the University of California, Irvine (UCI), are chosen to evaluate the accuracy of the various hybrid features selection models. In this paper, the procedures that are part of the proposed approaches are described and then evaluated for their performances.

scholarly journals Color space influence on ANN skin lesion classification using statistics texture feature

2021 ◽  
Vol 44 (1) ◽  
pp. 53-62
Author(s):  
Felicia Anisoara Damian ◽  
Simona Moldovanu ◽  
Luminita Moraru
Keyword(s):  
Color Space ◽  
Back Propagation ◽  
Texture Feature ◽  
Texture Features ◽  
Skin Lesions ◽  
Color Spaces ◽  
Color Channel ◽  
Feed Forward Back Propagation ◽  
Rgb Color Space ◽  
Hidden Layer

This study aims to investigate the ability of an artificial neural network to differentiate between malign and benign skin lesions based on two statistics terms and for RGB (R red, G green, B blue) and YIQ (Y luminance, and I and Q chromatic differences) color spaces. The targeted statistics texture features are skewness (S) and kurtosis (K) which are extracted from the histograms of each color channel corresponding to the color spaces and for the two classes of lesions: nevi and melanomas. The extracted data is used to train the Feed-Forward Back Propagation Networks (FFBPNs). The number of neurons in the hidden layer varies: it can be 8, 16, 24, or 32. The results indicate skewness features computed for the red channel in the RGB color space as the best choice to reach the goal of our study. The reported result shows the advantages of monochrome channels representation for skin lesions diagnosis.

FRACTAL CHARACTERIZATION OF BPN WEIGHTS EVOLUTION

2004 ◽  
Vol 14 (02) ◽  
pp. 139-145 ◽  
Author(s):  
S. GUNASEKARAN ◽  
B. VENKATESH ◽  
B. S. D. SAGAR
Keyword(s):  
Fractal Dimension ◽  
Dimensional Analysis ◽  
Back Propagation ◽  
Training Data ◽  
Data Sets ◽  
Training Phase ◽  
Data Set ◽  
Training Methodology ◽  
Propagation Network

Training methodology of the Back Propagation Network (BPN) is well documented. One aspect of BPN that requires investigation is whether or not the BPN would get trained for a given training data set and architecture. In this paper the behavior of the BPN is analyzed during its training phase considering convergent and divergent training data sets. Evolution of the weights during the training phase was monitored for the purpose of analysis. The evolution of weights was plotted as return map and was characterized by means of fractal dimension. This fractal dimensional analysis of the weight evolution trajectories is used to provide a new insight to understand the behavior of BPN and dynamics in the evolution of weights.

scholarly journals Feed-Forward Back Propagation Network for the prediction of diabetic retinopathy disorder

2021 ◽  
Vol 44 (1) ◽  
pp. 67-74
Author(s):  
Luminita Moraru ◽  
Simona Moldovanu ◽  
Andreea-Monica (Lăzărescu) Dincă
Keyword(s):  
Transfer Function ◽  
Back Propagation ◽  
Retinal Vessel ◽  
Disease Classification ◽  
Diabetic Patients ◽  
Main Task ◽  
Feed Forward ◽  
Feed Forward Back Propagation ◽  
Hidden Layer ◽  
Propagation Network

Some retina disorders mainly involve some blocked blood clots so that, the retinal vessels change their structure, being unable to completely nourish the retina. For an accurate investigation of retina disorders, the extraction of the retinal vessel anatomical structures or lesions is the main task. This paper reports a combination of various features extracted from retinal images, that are further used to train a Feed-Forward Back Propagation Network (FFBPN) as a decision system. The main goal is determining the combination of the appropriate features for more accurate classification of healthy and diseased patients. To achieve this goal, 120 binary images covering both categories of patients that belong to the STARE (Structured Analysis of the Retina) database were analyzed. The input data are the number of ridges, bifurcation, and bridges for retinal vessel pattern recognition. The FFBPNs with 4, 8, 12, 16, and 20 neurons in the hidden layer are trained. The FFBNP architecture with 12 neurons in the hidden layer, using the tansig transfer function in the hidden layer and linear transfer function in the output layer provides the most appropriate model for retinopathy disease classification. The correlation between the number of ridges and bridges computed for healthy patients (as actual values) and the number of ridges and bridges for diabetic patients (as predicted values) provides the best result, a regression coefficient (R) of 0. 8575 and a mean-square error (MSE) of 0.00163.

Reliability Inference for Products of Weibull Distribution by Using Artificial Neural Network Formal Description

2006 ◽  
Vol 326-328 ◽  
pp. 573-576
Author(s):  
Yung Chung Chen ◽  
Pei Hsi Lee ◽  
Chien Ming Chen
Keyword(s):  
Weibull Distribution ◽  
Research Result ◽  
Back Propagation ◽  
Accelerated Life Testing ◽  
Learning Ability ◽  
Accelerated Life ◽  
Input Variables ◽  
Hidden Layer ◽  
Propagation Network ◽  
Self Learning

Back-propagation network (BPN) has the advantage of simulating a nonlinear system that is difficult to describe by a physical model. This study introduces a back-propagation network methodology to estimate the accelerated life reliability. The environmental stresses and failure times are chosen as the input variables. An optimum prediction system is acquired by adjusting the number of neurons in the hidden layer and the output layer of neural networks. For a numerical example, the developed BPN architecture is applied to real accelerated life testing data of the STNLCD modules which are distributed as a Weibull distribution. By the research result, we can have the conclusion that the BPN methodology is practical to make the reliability inference with the advantages of self-learning ability even without mathematics models.

Prediction of geomagnetic storms from solar wind data with the use of a neural network

1994 ◽  
Vol 12 (1) ◽  
pp. 19-24 ◽  
Author(s):  
H. Lundstedt ◽  
P. Wintoft
Keyword(s):  
Neural Network ◽  
Solar Wind ◽  
Geomagnetic Storms ◽  
Activity Index ◽  
Back Propagation ◽  
Recovery Phase ◽  
Feed Forward Neural Network ◽  
Error Back Propagation ◽  
Hidden Layer ◽  
Propagation Network

Abstract. An artificial feed-forward neural network with one hidden layer and error back-propagation learning is used to predict the geomagnetic activity index (Dst) one hour in advance. The Bz-component and ΣBz, the density, and the velocity of the solar wind are used as input to the network. The network is trained on data covering a total of 8700 h, extracted from the 25-year period from 1963 to 1987, taken from the NSSDC data base. The performance of the network is examined with test data, not included in the training set, which covers 386 h and includes four different storms. Whilst the network predicts the initial and main phase well, the recovery phase is not modelled correctly, implying that a single hidden layer error back-propagation network is not enough, if the measured Dst is not available instantaneously. The performance of the network is independent of whether the raw parameters are used, or the electric field and square root of the dynamical pressure.

Sign in / Sign up

Export Citation Format

Share Document