Reforming Architecture and Loss Function of Artificial Neural Networks in Binary Classification Problems

2020 ◽  
Author(s):  
Mohammad A. Dastgheib ◽  
Abolghasem A. Raie
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Loss Function ◽  
Binary Classification ◽  
Classification Problems ◽  
Artificial Neural

scholarly journals Methodology for Knowledge Extraction from Trained Artificial Neural Networks

2018 ◽  
Vol 21 ◽  
pp. 6-14
Author(s):  
Andrey Bondarenko ◽  
Ludmila Aleksejeva
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Artificial Neural Networks ◽  
Decision Tree ◽  
Binary Classification ◽  
Knowledge Extraction ◽  
Classification Problems ◽  
Critical Areas ◽  
Artificial Neural ◽  
Fully Connected

Artificial neural networks are widely spread models that outperform more basic, but explainable machine learning models like classification decision tree. Although their lack of explainability severely limits their area of application. All mission critical areas or law regulated areas (like European GDPR) require model to be explained. Explainability allows model validation for correctness and lack of bias. Thus methods for knowledge extraction from artificial neural networks have gained attention and development efforts. Current paper addresses this problem and describes knowledge extraction methodology which can be applied to classification problems. It is based on previous research and allows knowledge to be extracted from trained fully connected feed-forward artificial neural network, from radial basis function neural network and from hyper-polytope based classifier in the form of binary classification decision tree, elliptical rules and If-Then rules.

scholarly journals PREDICTION OF FINANCIAL DISTRESS: CASE OF MINING ENTERPRISES IN CZECH REPUBLIC

2020 ◽  
Vol 15 (1) ◽  
pp. 1-14
Author(s):  
Zuzana Rowland ◽  
Alla Kasych ◽  
Petr Suler
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Artificial Neural Networks ◽  
Czech Republic ◽  
Financial Stability ◽  
Grey Zone ◽  
Financial Health ◽  
Classification Problems ◽  
Mining Companies ◽  
Artificial Neural

The ability to predict a company's financial health is a challenge for many researchers and scientists. It is also a distracting topic, as many other new approaches to financial health predictions have emerged in recent years. In this paper, we focused on identifying the financial health of mining companies in the Czech Republic. We chose the neural network method because, based on various instances of related research, neural networks represent a more reliable financial forecast than mathematical-statistical methods such as discriminant analysis and logistic regression. The concept of a neural network emerged with the first artificial neural networks, inspired by biological systems. The existence of prediction and classification problems directly predetermines artificial neural networks with respect to a given issue. We used the Amadeus database for processing, including financial indicators, SPSS, and Visual Gene Developer software. In total, we analyzed sixty-four mining companies. Complete data on financial stability were available for fifty-three companies, which we explored, and based on these results, identified financial situations for the other thirteen. Based on the available information, we processed a neural network and regression analysis. We managed to classify thirteen companies as solvent, insolvent, and in the grey zone, with the help of prediction.

scholarly journals MASKS: A Multi-Artificial Neural Networks System's verification approach

2020 ◽  
Author(s):  
Amirhoshang Hoseinpour Dehkordi ◽  
Majid Alizadeh ◽  
Ebrahim Ardeshir-Larijani ◽  
Ali Movaghar
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Artificial Neural Network ◽  
Artificial Neural Networks ◽  
Special Property ◽  
External Information ◽  
Multi Agent Systems ◽  
Classification Problems ◽  
Multi Agent ◽  
Artificial Neural

<div>Artificial Neural networks are one of the most widely applied approaches for classification problems. However, developing an errorless artificial neural network is in practice impossible, due to the statistical nature of such networks. The employment of artificial neural networks in critical applications has rendered any such emerging errors, in these systems, incredibly more significant. Nevertheless, the real consequences of such errors have not been addressed, especially due to lacking verification approaches. This study aims to develop a verification method that eliminates errors through the integration of multiple artificial neural networks. In order to do this, first of all, a special property has been defined, by the authors, to extract the knowledge of these artificial neural networks. </div><div>Furthermore, a multi-agent system has been designed, itself comprised of multiple artificial neural networks, in order to check whether the aforementioned special property has been satisfied, or not. Also, in order to help examine the reasoning concerning the aggregation of the distributed knowledge, itself gained through the combined effort of separate artificial neural networks and acquired external information sources, a dynamic epistemic logic-based method has been proposed.</div><div>Finally, we believe aggregated knowledge may lead to self-awareness for the system. As a result, our model shall be capable of verifying specific inputs, if the cumulative knowledge of the entire system proves its correctness. </div><div>In conclusion, and formulated for multi-agent systems, a knowledge-sharing algorithm (Abbr. MASKS) has been developed. Which after being applied on the MNIST dataset successfully reduced the error rate to roughly one-eighth of previous runs on individual artificial neural network in the same model. </div>

scholarly journals An attempt to model the relationship between MMI attenuation and engineering ground-motion parameters using artificial neural networks and genetic algorithms

2010 ◽  
Vol 10 (12) ◽  
pp. 2527-2537 ◽  
Author(s):  
G-A. Tselentis ◽  
L. Vladutu
Keyword(s):  
Neural Networks ◽  
Genetic Algorithms ◽  
Artificial Neural Networks ◽  
Ground Motion ◽  
Classification Problems ◽  
Attenuation Relations ◽  
Ground Motion Parameters ◽  
Ground Motion Attenuation ◽  
Motion Parameters ◽  
Artificial Neural

Abstract. Complex application domains involve difficult pattern classification problems. This paper introduces a model of MMI attenuation and its dependence on engineering ground motion parameters based on artificial neural networks (ANNs) and genetic algorithms (GAs). The ultimate goal of this investigation is to evaluate the target-region applicability of ground-motion attenuation relations developed for a host region based on training an ANN using the seismic patterns of the host region. This ANN learning is based on supervised learning using existing data from past earthquakes. The combination of these two learning procedures (that is, GA and ANN) allows us to introduce a new method for pattern recognition in the context of seismological applications. The performance of this new GA-ANN regression method has been evaluated using a Greek seismological database with satisfactory results.

scholarly journals Follow-Up and Risk Assessment in Patients with Myocardial Infarction Using Artificial Neural Networks

Complexity ◽  
2017 ◽  
Vol 2017 ◽  
pp. 1-8 ◽  
Author(s):  
Tatjana Gligorijević ◽  
Zoran Ševarac ◽  
Branislav Milovanović ◽  
Vlado Đajić ◽  
Marija Zdravković ◽  
...  
Keyword(s):  
Myocardial Infarction ◽  
Neural Networks ◽  
Artificial Neural Networks ◽  
Baroreflex Sensitivity ◽  
Training Dataset ◽  
Survival Prediction ◽  
Classification Problems ◽  
Risk Of Death ◽  
Increased Risk ◽  
Artificial Neural

Artificial neural networks (ANNs) are machine learning technique, inspired by the principles found in biological neurons. This technique has been used for prediction and classification problems in many areas of medical signal processing. The aim of this paper was to identify individuals with high risk of death after acute myocardial infarction using ANN. A training dataset for ANN was 1705 consecutive patients who underwent 24-hour ECG monitoring, short ECG analysis, noninvasive beat-to-beat heart-rate variability, and baroreflex sensitivity that were followed for 3 years. The proposed neural network classifier showed good performance for survival prediction: 88% accuracy, 81% sensitivity, 93% specificity, 0.85 F-measure, and area under the curve value of 0.77. These findings support the theory that patients with high sympathetic activity (reduced baroreflex sensitivity) have an increased risk of mortality independent of other risk factors and that artificial neural networks can indicate the individuals with a higher risk.

Recent Methodology in Connectionist Systems

2011 ◽  
pp. 94-106
Author(s):  
Ana B. Porto Pazos ◽  
Alberto Alvarellos González ◽  
Alejandro Pazos Sierra
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Glial Cells ◽  
Brain Networks ◽  
Classification Problems ◽  
Biological Phenomena ◽  
Artificial Neurons ◽  
Artificial Neural ◽  
The Brain

The Artificial NeuroGlial Networks, which try to imitate the neuroglial brain networks, appeared in order to process the information by means of artificial systems based on biological phenomena. They are not only made of artificial neurons, like the artificial neural networks, but also they are made of elements which try to imitate glial cells. An important glial role related with the processing of the brain information has been recently discovered but, as the functioning of the biological neuroglial networks is not exactly known, it is necessary to test several and different possibilities for creating Artificial NeuroGlial Networks. This chapter shows the functioning methodology of the Artificial NeuroGlial Networks and the application of a possible implementation of artificial glia to classification problems.

Toward principled regularization of deep networks—From weight decay to feature contraction

2019 ◽  
Vol 4 (30) ◽  
pp. eaaw1329 ◽  
Author(s):  
Atsuto Maki
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Classification Problems ◽  
Weight Decay ◽  
Target Error ◽  
Deep Networks ◽  
Artificial Neural

Training deep artificial neural networks for classification problems may benefit from exploiting intrinsic class similarities by way of network regularization that compensates for a drawback in the commonly used target error.

Sign in / Sign up

Export Citation Format

Share Document