DSpace Repository

Irregular Step of Changing for Neural Network Data Sets Improves the Accuracy of Resistive Sensors Calculation

Show simple item record

dc.contributor.author PENIN, Alexandr
dc.contributor.author SIDORENKO, Anatolie
dc.date.accessioned 2023-11-14T07:02:10Z
dc.date.available 2023-11-14T07:02:10Z
dc.date.issued 2023
dc.identifier.citation PENIN, Alexandr, SIDORENKO, Anatolie. Irregular Step of Changing for Neural Network Data Sets Improves the Accuracy of Resistive Sensors Calculation. In: 6th International Conference on Nanotechnologies and Biomedical Engineering: proc. of ICNBME-2023, 20–23, 2023, Chisinau, vol. 2: Biomedical Engineering and New Technologies for Diagnosis, Treatment, and Rehabilitation, 2023, p. 150-159. ISBN 978-3-031-42781-7. e-ISBN 978-3-031-42782-4. en_US
dc.identifier.isbn 978-3-031-42781-7
dc.identifier.isbn 978-3-031-42782-4
dc.identifier.uri https://doi.org/10.1007/978-3-031-42782-4_17
dc.identifier.uri http://repository.utm.md/handle/5014/24782
dc.description Acces full text - https://doi.org/10.1007/978-3-031-42782-4_17 en_US
dc.description.abstract A linear multiport is considered as a model of multiwire communication lines with resistive sensors of physical quantities or as a model of the sensors themselves. The calculation of sensor resistance from measured input currents using a neural network as an approximation problem is investigated. To demonstrate such a problem, using the parameters of multiports with one and two sensor loads, the corresponding number of input currents for a particular set of loads is calculated in a given range of changes in their values. The input and target vectors are composed in this way. The dimension of the input vector is equal to the number of input currents. Numerical experiments were carried out in the MATLAB Deep Learning package for the feed-forward network. The trained model is further tested on the control data set to ensure the given computation accuracy conditionally for “all possible” load values. The data sets generation is carried out with both the traditionally constant and an irregular or variable step of change in values. For the irregular step, in the divided data into training, test and validation sets, an internal pattern is excluded and the network shows a greater ability to generalize. The same control data set shows a reduction in relative error for a series of numerical experiments. The repeatability of training results with preset value of relative error is introduced, as special index for quantitative assessment of training quality when comparing training results of neural networks with generated training data. The obtained results provide the basis for the study of both chains with a large number of loads, as well as other approximation and regression problems. en_US
dc.language.iso en en_US
dc.publisher Springer Nature Switzerland en_US
dc.rights Attribution-NonCommercial-NoDerivs 3.0 United States *
dc.rights.uri http://creativecommons.org/licenses/by-nc-nd/3.0/us/ *
dc.subject resistive sensors en_US
dc.subject neural networks en_US
dc.subject data set en_US
dc.subject relative errors en_US
dc.title Irregular Step of Changing for Neural Network Data Sets Improves the Accuracy of Resistive Sensors Calculation en_US
dc.type Article en_US


Files in this item

The following license files are associated with this item:

This item appears in the following Collection(s)

  • 2023
    6th International Conference on Nanotechnologies and Biomedical Engineering, September 20–23, 2023, Chisinau, Moldova

Show simple item record

Attribution-NonCommercial-NoDerivs 3.0 United States Except where otherwise noted, this item's license is described as Attribution-NonCommercial-NoDerivs 3.0 United States

Search DSpace


Advanced Search

Browse

My Account