Training sample dimensions impact on artificial neural network optimal structure

2013 ◽  
Author(s):  
V. Z. Manusov ◽  
I. S. Makarov ◽  
S. A. Dmitriev ◽  
S. A. Eroshenko
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Training Sample ◽  
Optimal Structure ◽  
Artificial Neural

An Artificial Neural Network Recognition Method for Single-Phase Earth Fault of Ultra-High Voltage Power Grid

2013 ◽  
Vol 333-335 ◽  
pp. 1659-1662
Author(s):  
Hai Wei Lu ◽  
Gang Wu ◽  
Chao Xiong
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Single Phase ◽  
Training Sample ◽  
Normal Operation ◽  
Recognition Method ◽  
Time Motion ◽  
Training Samples ◽  
Artificial Neural ◽  
Network Diagnosis

Fault diagnosis is very important to make the system return to normal operation quickly after an accident. This paper diagnoses the specific component failure and failure area when the real-time motion information of inputting protection and switch transferred to a trained artificial neural network model by building an artificial neural network diagnosis model of components such as transmission line, bus bar and transformer, training the artificial neural network through taking the failure rule which is found by the historic fault data as a training sample. This method has obvious advantages in the accuracy and speed of diagnosis compared with the previous artificial neural network and overcomes the shortcomings of the incompletion of training samples and not well dealing with the heuristic knowledge.

scholarly journals SOFTWARE IMPLEMENTATION OF THE ARTIFICIAL NEURAL NETWORK FOR VIRTUAL OBJECTS СONTROL

2018 ◽  
pp. 72-78 ◽  
Author(s):  
Yu. B. Popova ◽  
S. V. Yatsynovich
Keyword(s):  
Neural Network ◽  
Mathematical Modeling ◽  
Artificial Neural Network ◽  
High Speed ◽  
Back Propagation ◽  
Computer Game ◽  
Training Sample ◽  
Software Implementation ◽  
Virtual Objects ◽  
Artificial Neural

Artificial neural networks (ANN) are now widely used in control and forecasting problems. The purpose of this work is the implementation of an artificial neural network for virtual objects control in a computer game of football. To achieve this goal, it is necessary to solve a number of problems related to mathematical modeling of ANN, algorithmization and software implementation. The paper deals with the mathematical modeling of an artificial neural network by the method of back propagation of an error, the algorithms for calculating neurons and for teaching ANN are presented. The software implementation of the artificial neural network was performed in the JavaScript language using the Node. js library, which assumed the role of a server for managing the game process. Some functions of the Underscore. js library were used to work with data arrays. The training sample consisted of more than 1000 sets of inputs and outputs, reflecting all possible situations. The results of software implementation of an artificial neural network are described on the example of virtual players control for a computer game. The results of the work show that ANN with a sufficiently high speed in real time gives the necessary direction for the player’s movement. The use of an artificial neural network has reduced the use of CPU time, which is extremely important in problems where rapid decision making is required, because complex calculations and prediction algorithms can not always be invested in 20 ms, which is fraught with skipping moves and losses. The simulated artificial neural network and the implemented algorithm of its learning can be used to solve other problems, for which only new data of the surrounding world are needed.

scholarly journals SELECTION OF THE OPTIMAL STRUCTURE OF HIDDEN LAYERS OF THE ARTIFICIAL NEURAL NETWORK FOR ENERGY EFFICIENCY ANALYSIS

2021 ◽  
Vol 3 (1) ◽  
pp. 30-36
Author(s):  
A. G. Kazarian ◽  
◽  
V. M. Teslyuk ◽  
I. Ya. Kazymyra ◽  
◽  
...  
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Internal Structure ◽  
Layer Structure ◽  
Operating Time ◽  
Main Idea ◽  
Optimal Structure ◽  
Output Data ◽  
Artificial Neural ◽  
Selection Of

A method for optimal structure selection of hidden layers of the artificial neural network (ANN) is proposed. Its main idea is the practical application of several internal structures of ANN and further calculation of the error of each hidden layer structure using identical data sets for ANN training. The method is based on the alternate comparison of the expected result values and the actual results of the feedforward artificial neural networks with a different number of inner layers and a different number of neurons on each layer. The method afforces searching the optimal internal structure of ANN for usage in the development of "smart" house systems and for calculation of the optimal energy consumption level in accordance with current conditions, such as room temperature, presence of people, and time of the day. The usage of the presented method allows to reduce the time spent on choosing the effective structure of the artificial neural network at the initial stages of research and to pay more attention to the relationship between the input and output data, as well as to such important parameters of the ANN learning process, as a number of training iterations, minimal training error, etc. The software has been developed that allows to carry out the processes of training, testing, and obtaining the output results of the algorithm of the artificial neural network, such as the expected value of power consumption and operating time of each individual appliance. The disadvantage of the approach used in finding the optimal internal structure of the artificial neural network is that each subsequent structure is created on the basis of the most efficient of the previously created structures without analyzing other structures that showed worse results with fewer hidden layers. It was found that to improve the solution of this problem it is necessary to create a mechanism which will be based on the analysis of input data, output data, will analyze the internal relationships between parameters and will optimize the network structure at each stage using certain logical rules according to the results obtained in the previous step. It is established that this problem is a nonlinear programming problem that can be solved in the further development of this study.

scholarly journals Prediction of erosional rates for cohesive sediments in annular flume experiments using artificial neural networks

2018 ◽  
Vol 1 (2) ◽  
pp. 99-111
Author(s):  
Olya Skulovich ◽  
Caroline Ganal ◽  
Leonie K. Nüßer ◽  
Catrina Cofalla ◽  
Holger Schuettrumpf ◽  
...  
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Mean Squared Error ◽  
Training Sample ◽  
River Rhine ◽  
Chemical Parameters ◽  
Flume Experiments ◽  
Annular Flume ◽  
Physical Chemical ◽  
Artificial Neural

Abstract Artificial neural network is used to predict development of suspended sediment concentration in annular flume experiments on cohesive sediment erosion. Natural sediment for the experiments was taken from the River Rhine and subjected to a consecutive increase in the bed shear stress. The development of the suspended particulate matter (SPM) was measured and then utilized for artificial neural network training, validation, and testing, including independent testing on new data sets. Several network configurations were examined, in particular, with and without autoregressive input. Additionally, relative importance of auxiliary physical-chemical parameters was analyzed. Artificial neural network with autoregressive input showed very high precision in the SPM prediction for all independent test cases achieving average mean squared error 0.034 and regression value 0.998. It was found that for an abundant training sample, the SPM parameter itself is enough to obtain high quality prediction. At the same time, physical-chemical parameters may provide some improvement to the artificial neural network prediction in cases that comprise values unprecedented in the training sample.

scholarly journals Classification of brain tumors using the multilayer perceptron artificial neural network

2018 ◽  
Vol 16 (36) ◽  
pp. 190-198
Author(s):  
Raid Adnan Omar
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Brain Tumor ◽  
Multilayer Perceptron ◽  
Research Work ◽  
Test Sample ◽  
Training Sample ◽  
Statistical Characteristics ◽  
Correct Classification ◽  
Artificial Neural

Information from 54 Magnetic Resonance Imaging (MRI) brain tumor images (27 benign and 27 malignant) were collected and subjected to multilayer perceptron artificial neural network available on the well know software of IBM SPSS 17 (Statistical Package for the Social Sciences). After many attempts, automatic architecture was decided to be adopted in this research work. Thirteen shape and statistical characteristics of images were considered. The neural network revealed an 89.1 % of correct classification for the training sample and 100 % of correct classification for the test sample. The normalized importance of the considered characteristics showed that kurtosis accounted for 100 % which means that this variable has a substantial effect on how the network perform when predicting cases of brain tumor, contrast accounted for 64.3 %, correlation accounted for 56.7 %, and entropy accounted for 54.8 %. All remaining characteristics accounted for 21.3-46.8 % of normalized importance. The output of the neural networks showed that sensitivity and specificity were scored remarkably high level of probability as it approached % 96.

scholarly journals Neurocontrol methods in the context of development of technical solutions for transition to unmanned navigation

2021 ◽  
Vol 2061 (1) ◽  
pp. 012115
Author(s):  
A I Epikhin
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Control System ◽  
Electric Propulsion ◽  
Control Object ◽  
Training Sample ◽  
Surface Vessels ◽  
Artificial Neural ◽  
Artificial Neural Network Ann ◽  
Technical Solutions

Abstract The paper considers the features and prospects of using neurocontrol methods in the context of development of technical solutions for transition to unmanned merchant vessels. The paper suggests a non-iterative training based artificial neural network (ANN), which is based on the principles of “direct inverse control” to control the speed and motion of unmanned surface vessels. The model is identified, and the structure of an artificial neural network and the diagram of the automatic control system (ACS) of an unmanned vessel (UV) are considered on the example of an electric propulsion vessel. A series of computational experiments is carried out to obtain a sufficiently complete training sample. and the control law is presented. The principle of the control system for an unmanned vessel is considered based on a neural network. At the next stage of the study, focus is on the synthesis of the optimal control system for UV navigation. The problem of the fastest motion of a third-order control object from one point (with any initial speed) to another (at the end point the vessel stops and the speed is zero) is considered. Based on the results of a series of experiments with the UV model, the controller parameters that provide the best indicators of control quality were set in the MATLAB Simulink environment.

scholarly journals An optimal feed-forward artificial neural network model and a new empirical correlation for prediction of the relative viscosity of Al2O3-engine oil nanofluid

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Mohammad Hemmat Esfe ◽  
Davood Toghraie
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Volume Fraction ◽  
Laboratory Data ◽  
Relative Viscosity ◽  
Optimal Structure ◽  
Engine Oil ◽  
Marquardt Algorithm ◽  
Optimal Feed ◽  
Artificial Neural

AbstractThis study presents the design of an artificial neural network (ANN) to evaluate and predict the viscosity behavior of Al2O3/10W40 nanofluid at different temperatures, shear rates, and volume fraction of nanoparticles. Nanofluid viscosity ($${\mu }_{nf}$$ μ nf ) is evaluated at volume fractions ($$\varphi$$ φ =0.25% to 2%) and temperature range of 5 to 55 °C. For modeling by ANN, a multilayer perceptron (MLP) network with the Levenberg–Marquardt algorithm (LMA) is used. The main purpose of this study is to model and predict the $${\mu }_{nf}$$ μ nf of Al2O3/10W40 nanofluid through ANN, select the best ANN structure from the set of predicted structures and manage time and cost by predicting the ANN with the least error. To model the ANN, $$\varphi$$ φ , temperature, and shear rate are considered as input variables, and $${\mu }_{nf}$$ μ nf is considered as output variable. From 400 different ANN structures for Al2O3/10W40 nanofluid, the optimal structure consisting of two hidden layers with the optimal structure of 6 neurons in the first layer and 4 neurons in the second layer is selected. Finally, the R regression coefficient and the MSE are 0.995838 and 4.14469E−08 for the optimal structure, respectively. According to all data, the margin of deviation (MOD) is in the range of less than 2% < MOD < + 2%. Comparison of the three data sets, namely laboratory data, correlation output, and ANN output, shows that the ANN estimates laboratory data more accurately.

Sign in / Sign up

Export Citation Format

Share Document