PL
 |
EN
Preferencje help
Polish version English version Język
Widoczny [Schowaj] Abstrakt
Liczba wyników

Znaleziono wyników: 20

Liczba wyników na stronie
first rewind previous Strona / 1 next fast forward last
Wyniki wyszukiwania
Wyszukiwano:
w słowach kluczowych:  neural nets
help Sortuj według:

help Ogranicz wyniki do:
first rewind previous Strona / 1 next fast forward last
1
Content available remote Optimization and Evaluation of Calibration for Low-cost Air Quality Sensors: Supervised and Unsupervised Machine Learning Models
Zhivkov Petar
Annals of Computer Science and Information Systems
|
2021
|
Vol. 25
255--258
EN
With the advancement of air pollution management, low-cost sensors are increasingly being used in air quality monitoring, but the data quality of these sensors is still a major source of concern. In this paper, data from five air monitoring stations in Sofia were compared to data from fixed low-cost PM sensors. The values of atmospheric pressure from low-cost sensors and the effects of relative humidity were investigated. A two-step model was created to refine the calibration process for low-cost PM sensors. At first, we calibrated the sensors with five separate supervised machine learning models and then the ANNf inal model with anomaly detection completed the results. The ANN-final model improved the R2 values of the PM10 determined by low-cost sensors from 0.62 to 0.95 as compared to standard instruments. In conclusion, the two-step calibration model proved to be a positive solution to addressing low-cost sensor efficiency issues.
2
Content available remote In-Bed Person Monitoring Using Thermal Infrared Sensors
Josse Elias, Nerborg Amanda, Hernandez Diaz Kevin, Alonso-Fernandez Fernando
Annals of Computer Science and Information Systems
|
2021
|
Vol. 25
121--125
EN
Technological solutions involving cameras can contribute to safety in home and healthcare, but they pose privacy issues. We use a low-resolution infrared thermopile array sensor, which offers more privacy, to determine if the user is on the bed. Two datasets were captured, one under constant conditions, and a second one under different variations. We test three machine learning algorithms under 10-fold cross validation, with the highest accuracy in the main dataset being 99%. The results with variable data show a lower reliability under certain circumstances, highlighting the need of extra work to meet the challenge of variations in the environment.
3
Content available remote Evaluation of Neural Network Transformer Models for Named-Entity Recognition on Low-Resourced Languages
Hanslo Ridewaan
Annals of Computer Science and Information Systems
|
2021
|
Vol. 25
115--119
EN
In this paper, transformer models are used to evaluate ten low-resourced South African languages for NER. Further, these transformer models are compared to bi-LSTM-aux and CRF models. The transformer models have the highest F-score of 84%. This result is significant within the context of the study, as previous research could not achieve F-scores of 80%. However, the CRF and bi-LSTM-aux models remain top performers in sequence tagging. Transformer models are viable for low-resourced languages. Future research could improve upon these findings by implementing a linear-complexity recurrent transformer variant.
4
Content available remote Explorations into Deep Learning Text Architectures for Dense Image Captioning
Toshevska Martina, Stojanovska Frosina, Zdravevski Eftim, Lameski Petre, Gievska Sonja
Annals of Computer Science and Information Systems
|
2020
|
Vol. 21
129--136
EN
Image captioning is the process of generating a textual description that best fits the image scene. It is one of the most important tasks in computer vision and natural language processing and has the potential to improve many applications in robotics, assistive technologies, storytelling, medical imaging and more. This paper aims to analyse different encoder-decoder architectures for dense image caption generation while focusing on the text generation component. Already trained models for image feature generation are utilized with transfer learning. These features are used for describing the regions using three different models for text generation. We propose three deep learning architectures for generating one-sentence captions of Regions of Interest (RoIs). The proposed architectures reflect several ways of integrating features from images and text. The proposed models were evaluated and compared with several metrics for natural language generation.
5
Content available remote Quantitative impact of label noise on the quality of segmentation of brain tumors on MRI scans
Marcinkiewicz Michał, Mrukwa Grzegorz
Annals of Computer Science and Information Systems
|
2019
|
Vol. 18
61--65
EN
Over the last few years, deep learning has proven to be a great solution to many problems, such as image or text classification. Recently, deep learning-based solutions have outperformed humans on selected benchmark datasets, yielding a promising future for scientific and real-world applications. Training of deep learning models requires vast amounts of high quality data to achieve such supreme performance. In real-world scenarios, obtaining a large, coherent, and properly labeled dataset is a challenging task. This is especially true in medical applications, where high-quality data and annotations are scarce and the number of expert annotators is limited. In this paper, we investigate the impact of corrupted ground-truth masks on the performance of a neural network for a brain tumor segmentation task. Our findings suggest that a) the performance degrades about 8% less than it could be expected from simulations, b) a neural network learns the simulated biases of annotators, c) biases can be partially mitigated by using an inversely-biased dice loss function.
6
Content available remote Predicting blood glucose using an LSTM neural network
El Idrissi Touria, Idri Ali, Abnane Ibtissam, Bakkoury Zohra
Annals of Computer Science and Information Systems
|
2019
|
Vol. 18
35--41
EN
Diabetes self-management relies on the blood glucose prediction as it allows taking suitable actions to prevent low or high blood glucose level. In this paper, we propose a deep learning neural network model for blood glucose prediction. The model is a sequential one using a Long- Short-Term Memory (LSTM) layer with two fully connected layers. Several experiments were carried out over data of 10 diabetic patients to decide on the model's parameters in order to identify the best variant of the model. The performance of the proposed model measured in terms of root mean square error (RMSE) was compared with the ones of an existing LSTM model and an autoregressive (AR) model. The results show that our model is significantly more accurate; in fact, our LSTM model outperforms the existing LSTM model for all patients and outperforms the AR model in 9 over 10 patients, besides, the performance differences were assessed by thWilcoxon statistical test. Furthermore, the mean of the RMSE of our model was 12.38 mg/dl while it was 28.84 mg/dl and 50.69 mg/dl for AR and the existing LSTM respectively.
7
Content available remote Machine-learning at the service of plastic surgery: a case study evaluating facial attractiveness and emotions using R language
Štěpánek Lubomír, Měšťák Jan, Kasal Pavel
Annals of Computer Science and Information Systems
|
2019
|
Vol. 18
107--112
EN
Since the plastic surgery should consider that facial impression is always dependent on current facial emotion, it came to be verified how precise classification of facial images into sets of defined facial emotions is.
8
Content available MLP neural nets in design of technological process
Rojek I.
Advances in Manufacturing Science and Technology
|
2015
|
Vol. 39, nr 1
87--95
EN
This paper proposes MLP neural nets to improve technological process design. The first stage of research concerned the creation of models to selection of machine tools, the second stage pertained the creation of models to selection of tools and the third stage concerned the creation of models to selection of machining parameters. In addition, use of tools is forecasted at various time intervals. The models were created using Statsoft STATISTICA Data Miner. These models were compared in order to obtain the best selection. Based on the models, it is possible to create different scenarios of the design of technological process.
PL
W artykule przedstawiono opracowanie sieci neuronowych MLP w celu poprawy projektowania procesu technologicznego. Pierwszy etap dotyczył tworzenia modeli wyboru obrabiarek, drugi modeli wyboru narzędzi i trzeci tworzenia modeli do wyboru parametrów obróbki skrawaniem. Dodatkowo w opracowanych modelach uwzględniono prognozowanie użycia narzędzi w różnych przedziałach czasowych. Stosowano program Statsoft STATISTICA Data Miner. Prowadzono analizy wyników dla poszczególnych modeli i opracowano kryteria doboru. Stwierdzono, że wprowadzenie sieci neronowych umożliwia tworzenie różnych scenariuszy projektowania procesu technologicznego.
9
Content available remote Zastosowanie sieci neuronowych do modelowania deformacji górniczych
Gruszczyński W.
Przegląd Górniczy
|
2010
|
T. 66, nr 7-8
7--11
PL
W artykule przedstawiono wyniki badań nad zastosowaniem sieci neuronowych do modelowania deformacji powierzchni wywołanych podziemną eksploatacją górniczą. Sieci neuronowe zastosowano bezpośrednio przy modelowaniu obniżeń, zaś nachylenia i krzywizny obliczano z użyciem wzorów stosowanych przy opracowywaniu wyników pomiarów geodezyjnych. Doświadczenia obejmowały jedynie wartości wskaźników deformacji dla asymptotycznych stanów niecek obniżeniowych. Badania przeprowadzono w trzech etapach, starając się sukcesywnie zawężać możliwości dotyczące sposobu wstępnego przetworzenia danych, architektury sieci i procesu ich uczenia. Pierwsze dwa etapy badań wykorzystywały dane symulowane z użyciem modelu Knothego i dotyczyły wyłącznie obniżeń, trzeci zaś wykorzystywał dane z linii obserwacyjnych z rejonu Górnego Śląska. Uzyskane wyniki wskazują na przydatność sieci neuronowych do modelowania wskaźników deformacji. Opracowana metoda jest względnie prosta w zastosowaniu i możliwa do zautomatyzowania.
XX
Results of research on neural nets using for surface deformation modelling caused by underground mining exploitation. The neural nets have been directly used for modelling of subsidence, while slopes and curvatures have been calculated using formulas used at surveying results working out. Experiments included only values of deformation indices for asymptotic states of subsidence troughs. Research has been carried out in three stages, trying to successively narrow down possibilities referring to the way of data processing, network archite cture and teaching process. The first two stages of research used simulated data with use of Knothe model and exclusively referred to subsidence, while the third used the data from the observation lines from the Upper Silesian region. The obtained results show usefulness of neural nets for deformation indices modelling. The worked out method is relatively simple in use and possible for automation.
10
The conception of neuronal modelling of strategic game
Piech H.
Informatyka Teoretyczna i Stosowana
|
2008
|
R. 8, nr 13
7-15
EN
The proposition of neural conception of strategic game modelling (with classical structure) is the aim of our paper. We assume that game has sizes 2*n. The creation of project of neuronal net with minimum number of layers and the number of neurons in every layer [1], which guarantee the obtainment of game solution is the target of analysis. Appearing two (or at least one) fundamental strategies and estimation game value (what is called the solution of game) will be the main effect of system functioning [2,3]. Proposed solution is one from possible but not optimal. It concludes from possibility of utilization in solving the strategic game system characteristics richer with information about strategies [4].
PL
Celem pracy jest zaproponowanie koncepcji przedstawienia gry strategicznej o prostej strukturze, to jest o rozmiarach 2*n w konwencji neuronowej. Celem analizy jest stworzenie projektu sieci neuronowej o minimalnej liczbie warstw i liczbie neuronów w każdej warstwie, która zapewni uzyskanie rozwiązania gry. Efektem działania systemu będzie wyłonienie dwóch (lub jednej) podstawowych strategii oraz określenie wartości gry (co nazywane jest w opisie rozwiązaniem gry). Proponowane rozwiązanie jest jednym z możliwych, lecz nie optymalnym. Powodem takiej konstatacji jest chociażby możliwość wykorzystania w systemie rozwiązującym grę strategiczna. bogatszych w informację charakterystyk analizowanych strategii.
11
Content available remote Analysis of technological process on the basis of efficiency criterion
Krupińska B., Szewieczek D.
Journal of Achievements in Materials and Manufacturing Engineering
|
2006
|
Vol. 17, nr 1-2
421--424
EN
Purpose: Technological process is a basic determinant of correctness of industrial company’s functioning on the market. In this connection they should treat with the priority all activities connected with technology, technology management and controlling, that is with their continuous improvement. Design/methodology/approach: The created model made it possible to analyze the chosen technological processes for the sake of efficiency criteria, which describe the relationships: operation – material, operation – machine, operation – man, operation – technological parameters. Findings The in this thesis conducted analysis includes hypothetical technological processes of producing typical machine pieces. Within their scope also the nonmaterial parameters of technological process have been taken into account, which resulted from arts of applied samples and projecting of the technological process. Practical implications: One has worked out an application that allows to analyze the efficiency of technological process in aspect of nonmaterial values and has used neuronal nets to verify particle indicators of quality of a process operation. Indicators appointment makes it possible to evaluate the process efficiency, which can constitute an optimization basis of particular operation. Originality/value: As a result of this analysis gained data enabled to optimize the technological process by estimating influence of the analyzed parameters on the whole of process and optimization conducting of any process.
12
Product quality control on the basis of stochastic neuronet model of through technology
Kuznetsov L., Domashnev P.
Systems Science
|
2005
|
Vol. 31, no 3
111-119
EN
It is convenient to use the mathematical apparatus of neuronic networks for simulation of composite production engineering. Neuronet model of technology is a superposition of neuron models of separate stages of process and represents correlation between technological parameters and production properties. Such a model is determined. On the basis of this model it is possible to obtain only the dot forecast of prospective properties of production, which can considerably differ from actual ones because of the error of input data. The paper deals with the ways of taking into account a stochasticity of technology and noise pollution of input data. This allows us to obtain confidence intervals for prospective values of production properties. Using stochastic neuronet model allows us to calculate control actions, which one will put to wishful values of the properties of production with an established risk level.
13
Calculation of PUMA 560 robot model using neural net
Lesewed A., Kurek J.
Elektronika : konstrukcje, technologie, zastosowania
|
2004
|
Vol. 45, nr 8-9
52-54
EN
The paper describes applications of feed-forward neural network for calculation of mathematical model for PUMA 560 robol. The model is based on the Lagrange-Euler formulation and described by a set of nonlinear differential and algebraic equations. The comparison of neural model and robot has been demonstrated.
PL
W pracy przedstawiono sposób wykorzystania sieci neuronowej feed-forward do wyznaczania matematycznego modelu robota PUMA 560. Model oparty na równaniach Lagrangea-Eulera zawiera nieliniowe różniczkowe i algebraiczne algorytmy. Przedstawiono także porównanie wyznaczonego modelu neuronowego i robota.
14
Content available remote Rüschendorf, Adaptive estimation of hazard functions
Döhler S., Rüschendorf L.
Probability and Mathematical Statistics
|
2002
|
Vol. 22, Fasc. 2
355--379
EN
In this paper we obtain convergence rates for sieved maximum-likelihood estimators of the log-hazard function in a censoring model. We also establish convergence results for an adaptive version of the estimator based on the method of structural risk-minimization. Applications are discussed to tensor product spline estimators as well as to neural net and radial basis function sieves. We obtain simplified bounds in comparison to the known literature. This allows us to derive several new classes of estimators and to obtain improved estimation rates. Our results extend to a more general class of estimation problems and estimation methods (minimum contrast estimators).
15
Content available remote Koszty jakości jako narzędzie zarządzania innowacjami w przedsiębiorstwie
Gawlik J., Boratyńska-Sala A.
Inżynieria Maszyn
|
2002
|
R. 7, z. 3
48-62
PL
W artykule przedstawiono pojęcia dotyczące innowacji w działalności przedsiębiorstwa i ich wpływ na jego funkcjonowanie. Omówiono koszty jakości ze szczególnym uwzględnieniem relacji producent-klient. Zaproponowano nowe podejście do analizy innowacyjnych działań pro-jakościowych, z wykorzystaniem metody trzech zbiorów (MTZ). Do wyznaczania kosztów zastosowano sieci neuronowe i arkusz kalkulacyjny Excel.
EN
In this report, notions concerning the innovations in the activities of an enterprise and their influence on its life have been shown. Quality costs, with special regard to relations producer - client have been discussed. A new approach to the analysis of pro - quality innovation activities with the three gatherings method (TGM) has been proposed. For the calculation of costs, neural nets and Excel calculating chart have been used.
16
Sterowanie „neuro-fuzzy” maszyną papierniczą
Nowacki Z.
Przegląd Papierniczy
|
2001
|
R. 57, nr 2
91--96
PL
W artykule wyjaśniono zasady sterowania rozmytego oraz budowy i działania sieci neuronowych. Podano również zarys zastosowania powyższych technik do sterowania maszyną papierniczą.
EN
In the paper the principles of the fuzzy logic control as well as of the design and operation of the neural network are explained. An outline of application of the above technique to the control of a paper machine is also presented.
17
Content available remote Parametric bivariate surfaces with neural networks
Mingo L. F., Martinez A., Gimenez V., Castellanos J.
Machine Graphics and Vision
|
2000
|
Vol. 9, No. 1/2
41-46
EN
This paper presents the application of Enhaced Neural Networks (ENN) to the field of Image Processing, more precisely, to the field of surfaces approximation via the generalization property that ENNs have. This architecture can perform a polynominal approximation of a given pattern sen in such a way that if the net has "n" hidden layers, then it will compute the "n"+2 degree polynominal approximation to its pattern set. Moreover, the behaviour of this net can be modified just modifying the activation functon f(x) of some neurons, in such a way that the net will compute the approximation to the pattern set using a function basis of functions f(x), this way the net computes the non lineal combination of basis elements to output the desired approximation. ENNs are used to represented a surface approximation. Some examples, concerning results when learning surfaces, are explained along this paper. Results are good since the Mean squared Error is very low and the computation time too.
18
Content available remote Selforganizing neural and neurofuzzy networks - a comparative study
Osowski S., Linh T.H.
Archives of Electrical Engineering
|
1999
|
Vol. 48, nr 1-2
199-217
EN
The paper presents the comparative analysis of neural and neurofuzzy networks in the aspect of learning algorithms and application features. Two kinds of networks have been compared: the hard clustering selforganizing feedforward structure and the structure of neurons of distributed activities, known as neurofuzzy network. The characteristic features of network structures as well as learning algorithms are analysed and compared. It was shown that neurofuzzy networks present some generalization of the classical networks and fulfill the same role. However their accuracy of representation of the data is much better at the same number of neurons.
PL
Artykuł przedstawia analizę porównawczą dwu rodzajów sieci neuronowych samoorganizujących się przez konkurencję: sieci klasycznych oraz sieci wykorzystujących zbiory rozmyte. Porównane są algorytmy uczenia obu sieci oraz ich najważniejsze cechy decydujące o zastosowaniach praktycznych. Zostało pokazane, że sieci rozmyte stanowią pewne udoskonalenie sieci klasycznych, pozwalające uzyskać lepsze wyniki odwzorowania danych uczących. Algorytmy uczące sieci rozmytych bazujących na współzawodnictwie są równie skuteczne a nawet prostsze niż sieci klasycznych. Z punktu widzenia zastosowan praktycznych oba rodzaje sieci spełniają podobną rolę i mogą wykonywać podobne zadania, choć sieci rozmyte wydają się lepiej przybliżać rzeczywiste uwarunkowania występujące w technice.
19
Content available remote Artificial neural networks in electrical drives control - a survey
Orłowska-Kowalska T.
Archives of Electrical Engineering
|
1998
|
Vol. 47, nr 2
127-143
EN
Actual trends of the artificial neural networks (ANN) applications in the modern electrical drive systems are presented in the paper. The possibilities of ANN applications for modelling, identification, state variables estimation and control, especially for adaptive control of DC and AC motors as well as for sensorless drives are demonstrated. The chosen examples of ANN applications in electrical drives are presented. The paper is subdivided into three main sections which describe, respectively, the principles of neural network technique, the possibility of ANN application in electrical drive modelling, estimation and control problems. The theoretical part of each topic is directly addressed to the application in electrical drives. In the last section of the paper some examples of ANN practical application in power electronics and motion control are presented, taken from the literature published in the last years.
PL
W artykule przedstawiono aktualne trendy związane z zastosowaniem sztucznych sieci neuronowych w napędzie elektrycznym. Opisano możliwości zastosowania sieci neuronowych do modelowania, identyfikacji, estymacji zmiennych stanu i sterowania silników elektrycznych prądu stałego i przemiennego, ze szczególnym uwzględnieniem problematyki sterowania adaptacyjnego i napędów bezczujnikowych. Każda z teoretycznych części artykułu znajduje swoje odniesienie do aplikacji w dziedzinie napędu elektrycznego. W ostatniej części przedstawiono wybrane przykłady praktycznych zastosowań sieci neuronowych w energoelektronice i automatyce napędu, zaczerpnięte z literatury opublikowanej w ostatnich latach.
20
Content available remote Artificial neural network approach to mixed boundary conditions identification
Ratajewicz-Mikołajczak E.
Archives of Electrical Engineering
|
1998
|
Vol. 47, nr 1
3-11
EN
The electric field model of a copper electrodeposition process on printed circuit board (PCB) by means of ultrasonic field is presented in this paper. The problems of copper plating, the current distribution in through-holes and ultrasonic intensification of the electrodeposition process are the subject of many investigations. The result of our measurement experiment indicates that a copper layer in a through-hole grows faster than on the flat surface of PCB. On the basis of measurements, the numerical (FEM) model has been proposed. The inverse problem have been formulated for those two coupled fields (by boundary conditions), making us possible to control the current density and the layer thickness. To solve this problem the artificial neural network (ANN) has been applied.
PL
Praca dotyczy modelowania procesu galwanicznego osadzania miedzi w polu ultradźwiękowym na płytkach obwodów drukowanych. Zagadnienia związane z osadzaniem miedzi wewnątrz wąskich otworów przelotowych i związane z tym trudności otrzymania warstw odpowiedniej jakości są wciąż przedmiotem badań. Zastosowanie pola ultradźwiękowego w tych procesach ma na celu przyspieszenie osadzania oraz poprawę jakości warstw. Przeprowadzone na modelu fizycznym płytki obwodu drukowanego doświadczenia wykazały, że w obecności pola ultradźwiękowego grubość otrzymanej w otworze warstwy jest większa niż na płaskiej powierzchni płytki. Na podstawie wyników pomiarów został stworzony model numeryczny układu. Dla pola elektrycznego procesu elektrolizy sprzężonego poprzez warunki brzegowe z polem ultradźdwiękowym sformułowano zagadnienie odwrotne. Rozwiązanie tego problemu daje możliwość kontroli rozkładu gęstości prądu na granicy faz metal - elektrolit, a tym samym grubości nanoszonej warstwy. Problem rozwiązano stosując metodę sztucznych sieci neuronowych.
first rewind previous Strona / 1 next fast forward last
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.