Search Results

1 - 10 of 51 items

  • Keyword: Artificial neural network x
Clear All Modify Search
Aspects Regarding The Possibility To Use “Neural Networks” In The Selection Of The “R & D” Strategy In The “Nonconventional Technologies” Field

Abstract

The paper referes to the possibility of using neural networks in selecting the “Research-Development” strategy, in the nonconventional technologies field. It presents a selection and the main key elements of the Research and Development (R & D) strategies, applicable in the nonconventional technologies (NT) field and, thus, the specific analytical elements of such a methodology. It also refers to the possibility of using neural networks (“NN”) up to the level of taking managerial decisions regarding the manufacturing processes. Research Objectives: Defining components from the neuron ‘s structure into the organizational systems, in order to select ptime strategies for the organizational management. Expected Results: Transposing the entire methodology in a software.

Open access
A nonlinear model for diagnosing malignancy in patients with exudative plural effusion using routine plural fluid findings

Abstract

Background: There is a challenge in diagnosing cancer in patients with exudative plural effusion using a noninvasive and accurate method.

Objective: We developed artificial neural network (ANN), as a nonlinear model, to discriminate malignant exudative plural effusion from nonmalignant based on routine pleural fluid findings.

Methods: The plural fluid parameters including total and differential cell counts, total proteins, lactate dehydrogenase (LDH), glucose, adenosine deaminase (ADA), as well as age and sex of 114 patients with exudative plural effusion were applied by models as input. The output was supposed to be the presence or absence of the cancer.

Results: The accuracy, sensitivity and specificity of ANN for predicting malignancy were 89.7%, 86.7%, and 91.7%, respectively. In addition, the neural network significantly outperformed the logistic regression model, as a linear model, (AUC: 0.892 vs. 0.633, respectively, p < 0.001).

Conclusion: The ANN is a novel accurate and noninvasive method that can be used clinically to diagnose malignancy in patients with exudative plural effusion.

Open access
Formation-based modelling and simulation of success in soccer

Abstract

The players’ positions of tactical groups in soccer can be mapped to formation-patterns by means of artificial neural networks (Kohonen, 1995). This way, the hundreds of positional situations of one half of a match can be reduced to about 20 to 30 types of formations (Grunz, Perl & Memmert, 2012; Perl, 2015), the coincidences of which can be used for describing and simulating tactical processes of the teams (Memmert, Lemmink & Sampaio, 2017): Developing and changing formations in the interaction with the opponent activities can be understood as a tactical game in the success context of ball control, space control and finally generating dangerous situations. As such it can be simulated using mathematical approaches like Monte Carlo-simulation and game theory in order to generate optimal strategic patterns. However, in accordance with results from game theory it turns out that in most cases the one optimal strategy does not exist (e.g. see Durlauf & Blume, 2010). Instead, a variety of partial strategies with different frequencies were necessary – an approach that is mathematically interesting but has nothing to do with soccer reality. An alternative approach, which is developed in the following, is to interrupt the strictness of a single strategic concept by creative elements, which improves flexible response to opponent activities as well as prevents from being analyzed by the opponent team.

The results of respective simulation reach from improving strategic behaviour to recognizing strategic patterns and in particular to analyzing role and meaning of creative elements.

Open access
Metamodelling of Inventory-Control Simulations Based on a Multilayer Perceptron

Abstract

Inventory control problems arise in various industries, and each single real-world inventory is replete with non-standard factors and subtleties. Practical stochastic inventory control problems are often analytically intractable, because of their complexity. In this regard, simulation-optimization is becoming more and more popular tool for solving complicated business-driven problems. Unfortunately, simulation, especially detailed, is both time and memory consuming. In the light of this fact, it may be more reasonable to use an alternative cheaper-to-compute metamodel, which is specifically designed in order to approximate an original simulation. In this research we discus metamodelling of stochastic multiproduct inventory control system with perishable products using a multilayer perceptron with a rectified linear unit as an activation function.

Open access
An ARMA Type Pi-Sigma Artificial Neural Network for Nonlinear Time Series Forecasting

Abstract

Real-life time series have complex and non-linear structures. Artificial Neural Networks have been frequently used in the literature to analyze non-linear time series. High order artificial neural networks, in view of other artificial neural network types, are more adaptable to the data because of their expandable model order. In this paper, a new recurrent architecture for Pi-Sigma artificial neural networks is proposed. A learning algorithm based on particle swarm optimization is also used as a tool for the training of the proposed neural network. The proposed new high order artificial neural network is applied to three real life time series data and also a simulation study is performed for Istanbul Stock Exchange data set.

Open access
Pattern Layer Reduction for a Generalized Regression Neural Network by Using a Self–Organizing Map

Abstract

In a general regression neural network (GRNN), the number of neurons in the pattern layer is proportional to the number of training samples in the dataset. The use of a GRNN in applications that have relatively large datasets becomes troublesome due to the architecture and speed required. The great number of neurons in the pattern layer requires a substantial increase in memory usage and causes a substantial decrease in calculation speed. Therefore, there is a strong need for pattern layer size reduction. In this study, a self-organizing map (SOM) structure is introduced as a pre-processor for the GRNN. First, an SOM is generated for the training dataset. Second, each training record is labelled with the most similar map unit. Lastly, when a new test record is applied to the network, the most similar map units are detected, and the training data that have the same labels as the detected units are fed into the network instead of the entire training dataset. This scheme enables a considerable reduction in the pattern layer size. The proposed hybrid model was evaluated by using fifteen benchmark test functions and eight different UCI datasets. According to the simulation results, the proposed model significantly simplifies the GRNN’s structure without any performance loss.

Open access
Theoretical Investigations for the Verification of Shear Centre and Deflection of Sigma Section by Back Propagation Neural Network Using Python

Abstract

The most important challenges in the construction field is to do the experimentation of the designing at real time. It leads to the wastage of the materials and time consuming process. In this paper, an artificial neural network based model for the verification of sigma section characteristics like shear centre and deflection are designed and verified. The physical properties like weight, depth, flange, lip, outer web, thickness, and area to bring shear centre are used in the model. Similarly, weight, purlin centres with allowable loading of different values used in the model for deflection verification. The overall average error rate as 1.278 percent to the shear centre and 2.967 percent to the deflection are achieved by the model successfully. The proposed model will act as supportive tool to the steel roof constructors, engineers, and designers who are involved in construction as well as in the section fabricators industry.

Open access
Planning Training Loads for The 400 M Hurdles in Three-Month Mesocycles Using Artificial Neural Networks

Abstract

This paper presents a novel approach to planning training loads in hurdling using artificial neural networks. The neural models performed the task of generating loads for athletes’ training for the 400 meters hurdles. All the models were calculated based on the training data of 21 Polish National Team hurdlers, aged 22.25 ± 1.96, competing between 1989 and 2012. The analysis included 144 training plans that represented different stages in the annual training cycle. The main contribution of this paper is to develop neural models for planning training loads for the entire career of a typical hurdler. In the models, 29 variables were used, where four characterized the runner and 25 described the training process. Two artificial neural networks were used: a multi-layer perceptron and a network with radial basis functions. To assess the quality of the models, the leave-one-out cross-validation method was used in which the Normalized Root Mean Squared Error was calculated. The analysis shows that the method generating the smallest error was the radial basis function network with nine neurons in the hidden layer. Most of the calculated training loads demonstrated a non-linear relationship across the entire competitive period. The resulting model can be used as a tool to assist a coach in planning training loads during a selected training period.

Open access
Issues in Using Self-Organizing Maps in Human Movement and Sport Science

Abstract

Self-Organizing Maps (SOMs) are steadily more integrated as data-analysis tools in human movement and sport science. One of the issues limiting researchers’ confidence in their applications and conclusions concerns the (arbitrary) selection of training parameters, their effect on the quality of the SOM and the sensitivity of any subsequent analyses. In this paper, we demonstrate how quality and sensitivity may be examined to increase the validity of SOM-based data-analysis. For this purpose, we use two related data sets where the research question concerns coordination variability in a volleyball spike. SOMs are an attractive tool for analysing this problem because of their ability to reduce the highdimensional time series to a two-dimensional problem while preserving the topological, non-linear relations in the original data. In a first step, we systematically search the SOM parameter space for a set of options that produces significantly lower continuity, accuracy and combined map errors and we discuss the sensitivity of SOM-based analyses of coordination variability to changes in training parameters. In a second step, we further investigate the effect of using different numbers of trials and variables on the SOM quality and sensitivity. These sensitivity analyses are able to validate the conclusions from statistical tests. Using this type of analysis can guide researchers to select SOM parameters that optimally represent their data and to examine how they affect the subsequent analyses. This may also enforce confidence in any conclusions that are drawn from studies using SOMs and enhance their integration in human movement and sport science.

Open access
Application of artificial neural networks to predict the deflections of reinforced concrete beams

Abstract

Nonlinear structural mechanics should be taken into account in the practical design of reinforced concrete structures. Cracking is one of the major sources of nonlinearity. Description of deflection of reinforced concrete elements is a computational problem, mainly because of the difficulties in modelling the nonlinear stress-strain relationship of concrete and steel. In design practise, in accordance with technical rules (e.g., Eurocode 2), a simplified approach for reinforced concrete is used, but the results of simplified calculations differ from the results of experimental studies.

Artificial neural network is a versatile modelling tool capable of making predictions of values that are difficult to obtain in numerical analysis. This paper describes the creation and operation of a neural network for making predictions of deflections of reinforced concrete beams at different load levels. In order to obtain a database of results, that is necessary for training and testing the neural network, a research on measurement of deflections in reinforced concrete beams was conducted by the authors in the Certified Research Laboratory of the Building Engineering Institute at Wrocław University of Science and Technology. The use of artificial neural networks is an innovation and an alternative to traditional methods of solving the problem of calculating the deflections of reinforced concrete elements. The results show the effectiveness of using artificial neural network for predicting the deflection of reinforced concrete beams, compared with the results of calculations conducted in accordance with Eurocode 2. The neural network model presented in this paper can acquire new data and be used for further analysis, with availability of more research results.

Open access