Ograniczanie wyników
Czasopisma help
Autorzy help
Lata help
Preferencje help
Widoczny [Schowaj] Abstrakt
Liczba wyników

Znaleziono wyników: 466

Liczba wyników na stronie
first rewind previous Strona / 24 next fast forward last
Wyniki wyszukiwania
Wyszukiwano:
w słowach kluczowych:  artificial neural network
help Sortuj według:

help Ogranicz wyniki do:
first rewind previous Strona / 24 next fast forward last
PL
W niniejszym artykule przedstawiono koncepcję i implementację modelu do rozpoznawania ras psów na podstawie zdjęcia. Do realizacji zadania wykorzystano model głębokiej sieci neuronowej bazujący na strukturze InceptionV3. Sieć została wytrenowana i przetestowana na zbiorze przypadków uczących liczącym ponad 20 tys. zdjęć 120 ras psów z zastosowaniem transferu wiedzy. Zbadano również wpływ jakości zdjęć na wyniki klasyfikacji. Sieć uzyskała bardzo dobre rezultaty zarówno w przypadku analizy typowych, jak i nietypowych zdjęć.
EN
This article presents the concept and implementation of a model for recognizing dog breeds based on an input image. The task was performed with the use of a deep neural network model based on the InceptionV3 structure. The neural network has been trained and tested on a dataset counting more than 20,000 images of 120 dog breeds using transfer learning technique. The impact of image quality on classification results was also examined. The model obtained very good results in the analysis of both typical and unusual input images.
EN
The paper deals with the problem of preliminary selection of pigmented lesionsfor further melanoma diagnosis. Several algorithms for input data pre-processingare proposed and artificial neural network for the examination of pigmented lesions is used. Computational results are reported.
EN
In modern computational science, the interplay existing between machine learning and optimization process marks the most vital developments. Optimization plays an important role in mechanical industries because it leads to reduce in material cost, time consumption and increase in production rate. The recent work focuses on performing the optimization task on Friction Stir Welding process for obtaining the maximum Ultimate Tensile Strength (UTS) of the friction stir welded joints. Two machine learning algorithms i.e. Artificial Neural Network (ANN) and Decision Trees regression model are selected for the purpose. The input variables are Tool Rotational Speed (RPM), Tool Traverse Speed (mm/min) and Axial Force (KN) while the output variable is Ultimate Tensile Strength (MPa). It is observed that in case of the Artificial Neural Networks the Root Mean Square Errors for training and testing sets are 0.842 and 0.808 respectively while in case of Decision Trees regression model, the training and testing sets result Root Mean Square Errors of 11.72 and 14.61. So, it can be concluded that ANN algorithm gives better and accurate result than Decision Tree regression algorithm.
PL
We współczesnych obliczeniach naukowych wzajemna zależność między uczeniem maszynowym a procesem optymalizacji wyznacza najbardziej istotne osiągnięcia. Optymalizacja odgrywa ważną rolę w przemyśle mechanicznym, ponieważ prowadzi do obniżenia kosztów materiałów, zużycia czasu i wzrostu szybkości produkcji. Ostatnie prace skupiają się na wykonaniu optymalizacji procesu zgrzewania tarciowego z przemieszaniem w celu uzyskania maksymalnej wytrzymałości na rozciąganie (UTS) połączeń zgrzewanych tarciowo z przemieszaniem. Do tego celu wybrano dwa algorytmy uczenia maszynowego, tj. Sztuczną sieć neuronową (ANN) i model decyzyjnego drzewa regresyjnego. Zmienne wejściowe to prędkość obrotowa narzędzia [obr/min], prędkość posuwu narzędzia [mm/min] i siła osiowa [kN], natomiast zmienną wyjściową jest maksymalna wytrzymałość na rozciąganie [MPa]. Zaobserwowano, że w przypadku sztucznych sieci neuronowych średnie błędy kwadratowe zbiorów uczących i testowych wynoszą odpowiednio 0,842 i 0,808, podczas gdy w przypadku modelu decyzyjnego drzewa regresji zbiory uczące i testujące dają średnie błędy kwadratowe 11,72 i 14,61. Można więc stwierdzić, że algorytm ANN daje lepsze i dokładniejsze wyniki niż algorytm regresji drzewa decyzyjnego.
EN
The paper deals with development neural network controller to ensure safe and reliable operation of damaged induction motor. It was chosen field-oriented control algorithm as a basis for neural controller aimed to provide reliable control signals both for healthy and damaged motor. It was investigated different compositions of neural network model. It was received simulation results for modified field-oriented control algorithm with neural network regulator, which showed similar results comparing to original model for healthy motor, which confirms correctness of developed model.
PL
Opisano zastosowanie sterownika wykorzystującego sieci neuronowe do stabilnej i niezawodnej pracy zdrowego lub uszkodzonego silnika indukcyjnego.
EN
The article presents a method enabling estimation of the selected power quality indicators at a given point of a power network, on the basis of the power quality indicators recorded at the nearest vicinity points. For needs of the estimations, artificial neural network algorithms were applied. The result is a neural model that defines the relationship between the power quality indicators of the same type, at adjacent points. The paper presents results of analyses and tests under real operating conditions of the distribution system.
PL
W artykule przedstawiono metodę umożliwiającą estymację wybranych wskaźników jakości energii elektrycznej w zadanym punkcie sieci elektroenergetycznej na podstawie wskaźników jakości energii elektrycznej zarejestrowanych w punktach leżących w najbliższym otoczeniu. Do estymacji wykorzystano algorytmy sztucznych sieci neuronowych. W rezultacie uzyskano neuronowy model określający relację pomiędzy wskaźnikami jakości energii elektrycznej tego samego typu w sąsiadujących ze sobą punktach. W artkule przedstawiono wyniki analiz i testów dla rzeczywistych warunków pracy sieci dystrybucyjnej.
EN
This article presents an analysis of the possibilities of using the pre-degraded GoogLeNet artificial neural network to classify inland vessels. Inland water authorities monitor the intensity of the vessels via CCTV. Such classification seems to be an improvement in their statutory tasks. The automatic classification of the inland vessels from video recording is a one of the main objectives of the Automatic Ship Recognition and Identification (SHREC) project. The image repository for the training purposes consists about 6,000 images of different categories of the vessels. Some images were gathered from internet websites, and some were collected by the project’s video cameras. The GoogLeNet network was trained and tested using 11 variants. These variants assumed modifications of image sets representing (e.g., change in the number of classes, change of class types, initial reconstruction of images, removal of images of insufficient quality). The final result of the classification quality was 83.6%. The newly obtained neural network can be an extension and a component of a comprehensive geoinformatics system for vessel recognition.
EN
This paper presents how machine learning techniques may be applied in the process of designing a compact dual-band H-shaped rectangular microstrip antenna (RMSA) operating in 0.75–2.20 GHz and 3.0–3.44 GHz frequency ranges. In the design process, the same dimensions of upper and lower notches are incorporated, with the centered position right in the middle. Notch length and width are verified for investigating the antenna. An artificial neural network (ANN) model is developed from the simulated dataset, and is used for shape prediction. The same dataset is used to create a mathematical model as well. The predicted outcome is compared and it is determined that the model relying on ANN offers better results.
8
Content available Sieci neuronowe w rozpoznawaniu pisma odręcznego
EN
Artificial neural networks consist of many simple elements capable of processing data. These are tools inspired by the construction of the human brain, used in machine learning. The aim of the research was to analyze the occuracy of the created neural network in the process of handwriting recognition. The article presents the results obtained during the learning and testing of a convolution network with a different number of hidden layers. Each time learning and testing the network was carried out using the same set of images (taken from the publicly available IAM database) depicting handwritten words in English.
PL
Sztuczne sieci neuronowe składają się z wielu prostych elementów zdolnych do przetwarzania danych. To narzędzia inspirowane budową ludzkiego mózgu, stosowane w uczeniu maszynowym. Celem badań była analiza dokładności odpowiedzi stworzonej sieci neuronowej w procesie rozpoznawania pisma odręcznego. W artykule przedstawiono wy-niki uzyskane podczas nauki i testowania sieci konwolucyjnej o różnej liczbie warstw ukrytych. Każdorazowo uczenie i testowanie sieci realizowane było za pomocą tego samego zbioru obrazów (zaczerpniętych z ogólnodostępnej bazy IAM Handrwiting Database) przedstawiających słowa pisane odręcznie w języku angielskim.
EN
This paper presents application of a neural network in the task of planning a mobile robot trajectory. First part contains a review of literature focused on the mobile robots’ orientation and overview of artificial neural networks’ application in area of robotics. In these sections devices and approaches for collecting data of mobile robots environment have been specified. In addition, the principle of operation and use of artificial neural networks in trajectory planning tasks was also presented. The second part focuses on the mobile robot that was designed in a 3D environment and printed with PLA material. The main onboard logical unit is Arduino Mega. Control system consist of 8-bits microcontrollers and 13 Mpix camera. Discussion in part three describes the system positioning capability using data from the accelerometer and magnetometer with overview of data filtration and the study of the artificial neural network implementation to recognize given trajectories. The last chapter contains a summary with conclusions.
EN
The article presents a model of operational fuel consumption by a passenger car from the B segment, powered by a spark ignition engine. The model was developed using artificial neural networks simulated in the Stuttgart Neural Network Simulator (SNNS) package. The data for the model was obtained from longterm operational tests, during which data from the engine control unit were recorded via the OBDII diagnostic interface. The model is based on neural networks with two hidden layers, the size of which was selected using an original iterative algorithm. During the structure selection process, a total of 576 different networks were tested. The analysis of the obtained test errors made it possible to select the optimal structure of the 6-19-17-1 model. The network input values were: vehicle speed and acceleration, road slope, throttle opening degree, selected gear number and engine speed. The networks were trained using the efficient RPROP method. A correctly trained network, based on the set parameters, was able to forecast the instantaneous fuel consumption. These forecasts showed a high correlation with the measured values. Average fuel consumption calculated on their basis was close to the real value, which was calculated on the basis of two consecutive fuelings of the vehicle.
PL
W artykule przedstawiono model eksploatacyjnego zużycia paliwa przez samochód osobowy z segmentu B, zasilany silnikiem o zapłonie iskrowym. Model opracowano przy wykorzystaniu sztucznych sieci neuronowych, których działanie symulowano w pakiecie Stuttgart Neural Network Simulator (SNNS). Dane do modelu pozyskano z długotrwałych badań eksploatacyjnych, podczas których rejestrowano przez interfejs diagnostyczny OBDII dane pochodzące z jednostki sterującej silnikiem. Model oparto na sieciach neuronowych o dwu warstwach ukrytych, których wielkość dobrano przy pomocy autorskiego, iteracyjnego algorytmu. Podczas procesu doboru struktury przebadano łącznie 576 różnych sieci. Analiza uzyskanych błędów testowania pozwoliła na wybór optymalnej struktury modelu 6-19-17-1. Wielkościami wejściowymi sieci były: prędkość i przyspieszenie pojazdu, nachylenie drogi, stopień otwarcia przepustnicy, numer wybranego biegu oraz prędkość obrotowa silnika. Sieci uczono przy użyciu wydajnej metody RPROP. Poprawnie nauczona sieć na podstawie zadanych parametrów była w stanie prognozować chwilowe zużycie paliwa. Prognozy te wykazywały wysoką korelację ze zmierzonymi wartościami. Obliczone na ich podstawie średnie zużycie paliwa było zbliżone do rzeczywistej wartości, którą obliczono na podstawie dwu kolejnych tankowań pojazdu.
EN
For designing a new energetic material with good performance, a knowledge of its density is required. In this study, the relationship between the densities of energetic co-crystals and their molecular structures was examined through a quantitative structure-property relationship (QSPR) method. The methodology of this research provides a new model which can relate the density of an energetic co-crystal to several molecular structural descriptors, which are calculated by Dragon software. It is indicated that the density of a co-crystal is a function of sp, OB, DU, nAT, as well as several non-additive structural parameters. The new recommended correlation was derived on the basis of the experimental densities of 50 co-crystals with various structures as a training set. The R2 or determination coefficient of the derived correlation was 0.937. This correlation provided a suitable estimate for a further 12 energetic co-crystals as a test set. Meanwhile, the predictive ability of the correlation was investigated through a cross validation method. Moreover, the new model has more reliability and performance for predicting the densities of energetic co-crystals compared to a previous one which was based on an artificial neural network approach. As a matter of fact, designing novel energetic co-crystals is possible by utilising the proposed method.
EN
Improving application efficiency is crucial for both the economic and environmental aspects of plant protection. Mathematical models can help in understanding the relationships between spray application parameters and efficiency, and reducing the negative impact on the environment. The effect of nozzle type, spray pressure, driving speed and spray angle on spray coverage on an artificial plant was studied. Artificial intelligence techniques were used for modeling and the optimization of application process efficiency. The experiments showed a significant effect of droplet size on the percent area coverage of the sprayed surfaces. A high value of the vertical transverse approach surface coverage results from coarse droplets, high driving speed, and nozzles angled forward. Increasing the vertical transverse leaving surface coverage, as well as the coverage of the sum of all sprayed surfaces, requires fine droplets, low driving speed, and nozzles angled backwards. The maximum coverage of the upper level surface is obtained with coarse droplets, low driving speed, and a spray angle perpendicular to the direction of movement. The choice of appropriate nozzle type and spray pressure is an important aspect of chemical crop protection. Higher upper level surface coverage is obtained when single flat fan nozzles are used, while twin nozzles produce better coverage of vertical surfaces. Adequate neural models and evolutionary algorithms can be used for pesticide application process efficiency optimization.
EN
Bone is a nonlinear, inhomogeneous and anisotropic material. To predict the behavior of bones expert systems are employed to reduce the computational cost and to enhance the accuracy of simulations. In this study, an artificial neural network (ANN) was used for the prediction of displacement in long bones followed by ex-vivo experiments. Three hydrated third metacarpal bones (MC3) from 3 thoroughbred horses were used in the experiments. A set of strain gauges were distributed around the midshaft of the bones. These bones were then loaded in compression in an MTS machine. The recordings of strains, load, load exposure time, and displacement were used as ANN input parameters. The ANN which was trained using 3,250 experimental data points from two bones predicted the displace-ment of the third bone (R2 ≥ 0.98). It was suggested that the ANN should be trained using noisy data points. The proposed modification in the training algorithm makes the ANN very robust against noisy inputs measurements. The performance of the ANN was evaluated in response to changes in the number of input data points and then by assuming a lack of strain data. A finite element analysis (FEA) was conducted to replicate one cycle of force-displace-ment experimental data (to gain the same accuracy produced by the ANN). The comparison of FEA and ANN displacement predictions indicates that the ANN produced a satisfactory outcome within a couple of seconds, while FEA required more than 160 times as long to solve the same model (CPU time: 5 h and 30 min).
EN
Diabetes mellitus (DM) is a multifactorial disease characterized by hyperglycemia. The type 1 and type 2 DM are two different conditions with insulin deficiency and insulin resistance, respectively. It may cause atherosclerosis, stroke, myocardial infarction and other relevant complications. It also features neurological degeneration with autonomic dysfunction to meet metabolic demand. The autonomic balance controls the physiological variables that exhibit nonlinear dynamics. Thus, in current work, nonlinear heart rate variability (HRV) parameters in prognosis of diabetes using artificial neural network (ANN) and support vector machine (SVM) have been demonstrated. The digital lead-I electrocardiogram (ECG) was recorded from male Wister rats of 10–12 week of age and 200 ± 20 gm of weight from control (n = 5) as well as from Streptozotocin induced diabetic rats (n = 5). A total of 526 datasets were computed from the recorded ECG data for evaluating thirteen nonlinear HRV parameters and used for training and testing of ANN. Using these parameters as inputs, the classification accuracy of 86.3% was obtained with an ANN architecture (13:7:1) at learning rate of 0.01. While relatively better accuracy of 90.5% was observed with SVM to differentiate the diabetic and control subjects. The obtained results suggested that nonlinear HRV parameters show distinct changes due to diabetes and hence along with machine learning tools, these can be used for development of noninvasive low-cost real-time prognostic system in predicting diabetes using machine learning techniques.
EN
Epilepsy is a widely spread neurological disorder caused due to the abnormal excessive neural activity which can be diagnosed by inspecting the electroencephalography (EEG) signals visually. The manual inspection of EEG signals is subjected to human error and is a tedious process. Further, an accurate diagnosis of generalized and focal epileptic seizures from normal EEG signals is vital for the supervision of pertinent treatment, life advancement of the subjects, and reduction in cost for the subjects. Hence the development of automatic detection of generalized and focal epileptic seizures from normal EEG signals is important. An approach based on tunable-Q wavelet transform (TQWT), entropies, Particle Swarm Optimization (PSO) and Artificial Neural Network (ANN) is proposed in this work for detection of epileptic seizures and its types. Two EEG databases namely, Karunya Institute of Technology and Sciences (KITS) EEG database and Temple University Hospital (TUH) database consisting of normal, generalized and focal EEG signals is used in this work to analyze the performance of the proposed approach. Initially, the EEG signals are decomposed into sub-bands using TQWT and the non-linear features like log energy entropy, Shannon entropy and Stein's unbiased risk estimate (SURE) entropy is computed from each sub-band. The informative features from the computed feature vectors are selected using PSO and fed into ANN for the classification of EEG signals. The proposed algorithm for KITS database achieved a maximum accuracy of 100% for four experimental cases namely, (i) normal-focal, (ii) normal-generalised, (iii) normal-focal + generalised and (iv) normal-focal-generalised. The TUH database achieved an accuracy of 95.1%, 97.4%, 96.2% and 88.8% for the four experimental cases. The proposed approach is promising and able to discriminate the epileptic seizure types with satisfactory classification performance.
EN
Predicting future blood glucose (BG) levels for diabetic patients will help them avoid potentially critical health issues. We demonstrate the use of machine learning models to predict future blood glucose levels given a history of blood glucose values as the single input parameter. We propose an Artificial Neural Network (ANN) model with time-domain attributes to predict blood glucose levels 15, 30, 45 and 60 min in the future. Initially, the model's features are selected based on the previous 30 min of BG measurements before a trained model is generated for each patient. These features are combined with time-domain attributes to give additional inputs to the proposed ANN. The prediction model was tested on 12 patients with Type 1 diabetes (T1D) and the results were compared with other data-driven models including the Support Vector Regression (SVR), K-Nearest Neighbor (KNN), C4.5 Decision Tree (DT), Random Forest (RF), Adaptive Boosting (AdaBoost) and eXtreme Gradient Boosting (XGBoost) models. Our results show that the proposed BG prediction model that is based on an ANN outperformed all other models with an average Root Mean Square Error (RMSE) of 2.82, 6.31, 10.65 and 15.33 mg/dL for Prediction Horizons (PHs) of 15, 30, 45 and 60 min, respectively. Our testing showed that combining time-domain attributes into the input data resulted in enhanced performance of majority of prediction models. The implementation of proposed prediction model allows patients to obtain future blood glucose levels, so that the preventive alerts can be generated before critical hypoglycemic/ hyperglycemic events occur.
EN
Solar air heater (SAH) is an important device for solar energy utilization which is used for space heating, crop drying, timber seasoning etc. Its performance mainly depends on system parameters, operating parameters and meteorological parameters. Many researchers have been used these parameters to predict the performance of SAH by analytical or conventional approach and artificial neural network (ANN) technique, but performance prediction of SAH by using relevant input parameters has not been done so far. Therefore, relevant input parameters have been considered in this study. Total ten parameters were used such as mass flow rate, ambient temperature, wind speed, relative humidity, fluid inlet temperature, fluid mean temperature, plate temperature, wind direction, solar elevation and solar intensity to find out the relevant parameters for ANN prediction. Seven different neural models have been constructed using these parameters. In each model 10 to 20 neurons have been selected to find out the optimal model. The optimal neural models for ANN-I, ANN-II, ANN-III, ANN-IV, ANN-V, ANN-VI and ANN-VII were obtained as 10-17-1, 8-14-1, 6-16-1, 5- 14-1, 4-17-1, 3-16-1 and 2-14-1, respectively. It has been found that ANN-II model with 8-14-1 is the optimal model as compared to other neural models. Values of the sum of squared errors, mean relative error, and coefficient of determination were found to be 0.02138, 1.82% and 0.99387, respectively, which shows that the ANN-II developed with mass flow rate, ambient temperature, inlet and mean temperature of air, plate temperature, wind speed and direction, relative humidity, and relevant input parameters performed better. The above results show that these eight parameters are relevant for prediction.
EN
Numerical predictions of heat transfer under laminar conditions in a square duct with ribs are presented in this paper. Ribs are provided on top and bottom walls in a square duct in a staggered manner. The flow rates have been varied between Reynolds number 200 and 600. Various configurations of ribs by varying length, width and depth have been investigated for their effect on heat transfer, friction factor and entropy augmentation generation number. Further artificial neural network integrated with genetic algorithm was used to minimize the entropy augmentation generation number (performance factor) by selecting the optimum rib dimensions in a selected range. Genetic algorithm is compared with microgenetic algorithm to examine the reduction in computational time for outlay of solution accuracy.
EN
This article discusses the results of studies using the developed artificial neural networks in the analysis of the occurrence of the four main mechanisms destroying the selected forging tools subjected to five different surface treatment variants (nitrided layer, pad welded layer and three hybrid layers, i.e. AlCrTiSiN, Cr/CrN and Cr/AlCrTiN). Knowledge of the forging tool durability, needed in the process of artificial neural network training, was included in the set of training data (about 800 records) derived from long-term comprehensive research carried out under industrial conditions. Based on this set, neural networks with different architectures were developed and the results concerning the intensity of the occurrence of thermal-mechanical fatigue, abrasive wear, mechanical fatigue and plastic deformation were generated for each type of the applied treatment relative to the number of forgings, pressure, friction path and temperature.
EN
Purpose: This article uses soft computing-based techniques to elaborate a study on the prediction of the friction angle of clay. Design/methodology/approach: A total of 30 data points were collected from the literature to predict the friction angle of the clay. To achieve the friction angle, the independent parameters sand content, silt content, plastic limit and liquid limit were used in the soft computing techniques such as artificial neural networks, M5P model tree and multi regression analysis. Findings: The major findings from this study are that the artificial neural networks are predicting the friction angle of the clay accurately than the M5P model and multi regression analysis. The sensitivity analysis reveals that the clay content is the major influencing independent parameter to predict the friction angle of the clay followed by sand content, liquid limit and plastic limit. Research limitations/implications: The proposed expressions can used to predict the friction angle of the clay accurately but can be further improved using large data for a wider range of applications. Practical implications: The proposed equations can be used to calculate the friction angle of the clay based on sand content, silt content, plastic limit and liquid limit. Originality/value: There is no such expression available in the literature based on soft computing techniques to calculate the friction angle of the clay.
first rewind previous Strona / 24 next fast forward last
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.