Informacja

Drogi użytkowniku, aplikacja do prawidłowego działania wymaga obsługi JavaScript. Proszę włącz obsługę JavaScript w Twojej przeglądarce.

English (EN)·Polski (PL)·Yкраїнський (UK)
Przejdź do strony domowej biblioteki Centralna Biblioteka Wojskowa Link otwiera się w nowym oknie Logo biblioteki Prolib Integro - strona głównaCentralna Biblioteka Wojskowa

Wyszukujesz frazę "approximation network" wg kryterium: Temat

  Lista filtrów
Wyrównaj
    Wyświetlanie 1-15 z 15
    Tytuł:
    Towards a very fast feedforward multilayer neural networks training algorithm
    Autorzy:
    Bilski, Jarosław
    Kowalczyk, Bartosz
    Kisiel-Dorohinicki, Marek
    Siwocha, Agnieszka
    Żurada, Jacek
    Powiązania:
    https://bibliotekanauki.pl/articles/2147135.pdf
    Data publikacji:
    2022
    Wydawca:
    Społeczna Akademia Nauk w Łodzi. Polskie Towarzystwo Sieci Neuronowych
    Tematy:
    neural network training algorithm
    QR decomposition
    scaled Givens rotation
    approximation
    classification
    Opis:
    This paper presents a novel fast algorithm for feedforward neural networks training. It is based on the Recursive Least Squares (RLS) method commonly used for designing adaptive filters. Besides, it utilizes two techniques of linear algebra, namely the orthogonal transformation method, called the Givens Rotations (GR), and the QR decomposition, creating the GQR (symbolically we write GR + QR = GQR) procedure for solving the normal equations in the weight update process. In this paper, a novel approach to the GQR algorithm is presented. The main idea revolves around reducing the computational cost of a single rotation by eliminating the square root calculation and reducing the number of multiplications. The proposed modification is based on the scaled version of the Givens rotations, denoted as SGQR. This modification is expected to bring a significant training time reduction comparing to the classic GQR algorithm. The paper begins with the introduction and the classic Givens rotation description. Then, the scaled rotation and its usage in the QR decomposition is discussed. The main section of the article presents the neural network training algorithm which utilizes scaled Givens rotations and QR decomposition in the weight update process. Next, the experiment results of the proposed algorithm are presented and discussed. The experiment utilizes several benchmarks combined with neural networks of various topologies. It is shown that the proposed algorithm outperforms several other commonly used methods, including well known Adam optimizer.
    Źródło:
    Journal of Artificial Intelligence and Soft Computing Research; 2022, 12, 3; 181--195
    2083-2567
    2449-6499
    Pojawia się w:
    Journal of Artificial Intelligence and Soft Computing Research
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Theory I: Deep networks and the curse of dimensionality
    Autorzy:
    Poggio, T.
    Liao, Q.
    Powiązania:
    https://bibliotekanauki.pl/articles/200623.pdf
    Data publikacji:
    2018
    Wydawca:
    Polska Akademia Nauk. Czytelnia Czasopism PAN
    Tematy:
    deep network
    shallow network
    convolutional neural network
    function approximation
    deep learning
    sieci neuronowe
    aproksymacja funkcji
    uczenie głębokie
    Opis:
    We review recent work characterizing the classes of functions for which deep learning can be exponentially better than shallow learning. Deep convolutional networks are a special case of these conditions, though weight sharing is not the main reason for their exponential advantage.
    Źródło:
    Bulletin of the Polish Academy of Sciences. Technical Sciences; 2018, 66, 6; 761-773
    0239-7528
    Pojawia się w:
    Bulletin of the Polish Academy of Sciences. Technical Sciences
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    The ANN approximation of the CH4 combustion model : the mixture composition
    Autorzy:
    Kowalski, J.
    Powiązania:
    https://bibliotekanauki.pl/articles/246942.pdf
    Data publikacji:
    2010
    Wydawca:
    Instytut Techniczny Wojsk Lotniczych
    Tematy:
    modeling
    internal combustion engines
    approximation
    artificial neural network
    combustion process
    chemical species
    Opis:
    The calculation of the changing of the combustion mixture composition during the combustion process of the CH4 is presented of the paper. Correct calculation results of the mixture composition during the combustion process in combustion chambers of internal combustion engines is important to define the heat release calculation, modeling and simulation of the combustion phenomena. The paper presents results of calculations for the GriMech 3 kinetic mechanism of the methane combustion for different thermodynamic parameters and the composition of the combusted mixture. Results of the kinetic calculation of combustion process are qualitatively consistent with the data available in literature. The second purpose of research was the approximation of obtained results with the trained artificial neural network. Input data needed to approximate mole fractions of considered in the GriMech 3 mechanism combustion process chemical species consisted of 52 mole fractions of initial chemical species and temperature and pressure process. For all considered chemical species the mean square error did not exceed a value of 1-10-2 %, but the maximum error for a single value of 43 species excess even more than 100% of the value of mole fraction values taken from kinetic calculations. Single values of errors disqualify the neural network application for modeling of mole fractions of chemical species.
    Źródło:
    Journal of KONES; 2010, 17, 2; 233-240
    1231-4005
    2354-0133
    Pojawia się w:
    Journal of KONES
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    The ANN approximation of the CH4 combustion model : the heat release
    Autorzy:
    Kowalski, J.
    Powiązania:
    https://bibliotekanauki.pl/articles/246946.pdf
    Data publikacji:
    2010
    Wydawca:
    Instytut Techniczny Wojsk Lotniczych
    Tematy:
    modelling
    internal combustion engines
    approximation
    artificial neural network
    combustion process
    heat release
    Opis:
    The calculation of the heat release from the combustion process of the CH4 is presented of the paper. Correct calculation results of the heat released from combustion is important for design, modelling and testing phenomena in combustion chambers of internal combustion engines. The paper presents results of calculations for the kinetic mechanism of methane combustion GriMech 3 for different thermodynamic parameters and composition of the combusted mixture. The calculations were performed for all possible configurations of the variable temperaturę range from 1100K to 3600K, the variable pressure in the range of 2MPa to 5MPa, variable humidity of charged air from 10 to 30 grams of water per l kg of air and variable mole fractions of charge air. Results of the kinetic calculation of combustion process are qualitatively consistent with the data available in literature. The next stage of research was approximation of obtained results with the trained artificial neural network. Input data needed to approximate the energy of the combustion process consisted of 52 mole fractions of chemical species and temperature and pressure process. Approximation results have meant square error not exceeded 0.04% for the test data and 0.02% for the validation data. The maximum error for a single result was 1.9% compared to data obtained with chemical kinetic calculations.
    Źródło:
    Journal of KONES; 2010, 17, 2; 225-232
    1231-4005
    2354-0133
    Pojawia się w:
    Journal of KONES
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Optimization of pin fin heat sink by application of CFD simulations and DOE methodology with neural network approximation
    Autorzy:
    Kasza, K.
    Malinowski, Ł.
    Królikowski, I.
    Powiązania:
    https://bibliotekanauki.pl/articles/265135.pdf
    Data publikacji:
    2013
    Wydawca:
    Uniwersytet Zielonogórski. Oficyna Wydawnicza
    Tematy:
    wymiana ciepła
    optymalizacja projektu
    radiator
    sieci neuronowe
    modelowanie numeryczne
    polimery
    heat transfer
    design optimization
    heat sink
    neural network approximation
    numerical modeling
    thermally conductive polymer
    Opis:
    A design optimization of a staggered pin fin heat sink made of a thermally conductive polymer is presented. The influence of several design parameters like the pin fin height, the diameter, or the number of pins on thermal efficiency of the natural convection heat sink is studied. A limited number of representative heat sink designs were selected by application of the design of experiments (DOE) methodology and their thermal efficiency was evaluated by application of the antecedently validated and verified numerical model. The obtained results were utilized for the development of a response surface and a typical polynomial model was replaced with a neural network approximation. The particle swarm optimization (PSO) algorithm was applied for the neural network training providing very accurate characterization of the heat sink type under consideration. The quasi-complete search of defined solution domain was then performed and the different heat sink designs were compared by means of thermal performance metrics, i.e., array, space claim and mass based heat transfer coefficients. The computational fluid dynamics (CFD) calculations were repeated for the most effective heat sink designs.
    Źródło:
    International Journal of Applied Mechanics and Engineering; 2013, 18, 2; 365-381
    1734-4492
    2353-9003
    Pojawia się w:
    International Journal of Applied Mechanics and Engineering
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Online learning algorithm for zero-sum games with integral reinforcement learning
    Autorzy:
    Vamvoudakis, K. G.
    Vrabie, D.
    Lewis, F. L.
    Powiązania:
    https://bibliotekanauki.pl/articles/91780.pdf
    Data publikacji:
    2011
    Wydawca:
    Społeczna Akademia Nauk w Łodzi. Polskie Towarzystwo Sieci Neuronowych
    Tematy:
    learning
    online algorithm
    zero-sum game
    game
    infinite horizon
    Hamilton-Jacobi-Isaacs equation
    approximation network
    optimal value function
    adaptive control tuning algorithm
    Nash solution
    Opis:
    In this paper we introduce an online algorithm that uses integral reinforcement knowledge for learning the continuous-time zero sum game solution for nonlinear systems with infinite horizon costs and partial knowledge of the system dynamics. This algorithm is a data based approach to the solution of the Hamilton-Jacobi-Isaacs equation and it does not require explicit knowledge on the system’s drift dynamics. A novel adaptive control algorithm is given that is based on policy iteration and implemented using an actor/ disturbance/critic structure having three adaptive approximator structures. All three approximation networks are adapted simultaneously. A persistence of excitation condition is required to guarantee convergence of the critic to the actual optimal value function. Novel adaptive control tuning algorithms are given for critic, disturbance and actor networks. The convergence to the Nash solution of the game is proven, and stability of the system is also guaranteed. Simulation examples support the theoretical result.
    Źródło:
    Journal of Artificial Intelligence and Soft Computing Research; 2011, 1, 4; 315-332
    2083-2567
    2449-6499
    Pojawia się w:
    Journal of Artificial Intelligence and Soft Computing Research
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Niektóre osobliwości aproksymacji neuronowej na przykładzie odwrotnego zadania kinematyki
    Some peculiarities of neural approximation on example of inverse kinematic problem
    Autorzy:
    Bartecki, K.
    Powiązania:
    https://bibliotekanauki.pl/articles/155167.pdf
    Data publikacji:
    2010
    Wydawca:
    Stowarzyszenie Inżynierów i Techników Mechaników Polskich
    Tematy:
    sztuczna sieć neuronowa
    aproksymacja funkcji
    odwrotne zadanie kinematyki
    artificial neural network
    function approximation
    inverse kinematics
    Opis:
    W artykule wskazano na pewne charakterystyczne aspekty związane z zastosowaniem jednokierunkowych sieci neuronowych jako uniwersalnych układów aproksymujących złożone zależności nieliniowe. Zaprezentowany przykład dotyczy klasycznego problemu z dziedziny robotyki -tzw. odwrotnego zadania kinematyki. Zademonstrowano wpływ właściwego doboru struktury sieci, jej algorytmu uczenia oraz wzorców uczących na jakość aproksymacji neuronowej.
    Characteristic features of feedforward artificial neural networks, acting as universal function approximators, are presented. The problem under consideration concerns inverse kinematics of a two-link planar manipulator (Fig. 1). As shown in this paper, a two-layer, feedforward neural network is able to learn the nonlinear mapping between the end effector position domain and the joint angle domain of the manipulator (Fig. 2). However, a necessary condition for achieving the required approximation quality is proper selection of the network structure, especially with respect to the number of nonlinear, sigmoidal units in its hidden layer. Using too few neurons in this layer results in underfitting (Fig. 3), while too many neurons bring the problem of overfitting (Figs 6 and 7). The effect of learning algorithm efficiency as well as proper choice of learning data set on the network performance is also demonstrated (Fig. 8). Apart from the general conclusions concerning neural approximation, the presented results show also the possibility of neural control of robotic manipulator trajectory.
    Źródło:
    Pomiary Automatyka Kontrola; 2010, R. 56, nr 6, 6; 589-592
    0032-4140
    Pojawia się w:
    Pomiary Automatyka Kontrola
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Investigation of G-networks with restart at a non-stationary mode and their application
    Autorzy:
    Matalytski, Mikhail
    Naumenko, Victor
    Kopats, Dmitry
    Powiązania:
    https://bibliotekanauki.pl/articles/122538.pdf
    Data publikacji:
    2019
    Wydawca:
    Politechnika Częstochowska. Wydawnictwo Politechniki Częstochowskiej
    Tematy:
    queueing network
    G-network
    positive and negative customers
    non-stationary regime
    multiple type customers
    signals
    restart
    nonstationary state probabilities
    successive approximation method
    sieć kolejkowa
    sieć G
    sygnały
    metoda aproksymacji
    aproksymacja
    teoria kolejek
    pozytywny klient
    negatywny klient
    Opis:
    This article discusses the question of restarting the script, when restart is used by many users of the information network, which can be modelled as a G-network. Negative claims simulate the crash of the script and the re-sending of the request. Investigation of an open queuing network (QN) with several types of positive customers with the phase type of distribution of their service time and one type of negative customers have been carried out. Negative customers are signals whose effect is to restart one customers in a queue. A technique is proposed for finding the probability of states. It is based on the use of a modified method of successive approximations, combined with the method of a series. The successive approximations converge with time to a stationary distribution of state probabilities, the form of which is indicated in the article, and the sequence of approximations converges to the solution of the difference-differential equations (DDE) system. The uniqueness of this solution is proved. Any successive approximation is representable in the form of a convergent power series with an infinite radius of convergence, the coefficients of which satisfy recurrence relations, which is convenient for computer calculations. A model example illustrating the finding of time-dependent probabilities of network states using the proposed technique is also presented.
    Źródło:
    Journal of Applied Mathematics and Computational Mechanics; 2019, 18, 2; 41-51
    2299-9965
    Pojawia się w:
    Journal of Applied Mathematics and Computational Mechanics
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Implementacja sztucznej sieci neuronowej w architekturze równoległej z wykorzystaniem protokołu MPI
    Parallel implementation of artificial neural network with use of MPI protocol
    Autorzy:
    Bartecki, K.
    Czorny, M.
    Powiązania:
    https://bibliotekanauki.pl/articles/153068.pdf
    Data publikacji:
    2011
    Wydawca:
    Stowarzyszenie Inżynierów i Techników Mechaników Polskich
    Tematy:
    sztuczna sieć neuronowa
    architektura równoległa
    aproksymacja funkcji
    artificial neural network
    parallel architecture
    function approximation
    Opis:
    W artykule wskazano na pewne aspekty związane z implementacją jednokierunkowej sieci neuronowej w architekturze równoległej z wykorzystaniem standardu przesyłania komunikatów MPI. Zaprezentowany przykład zastosowania sieci dotyczy klasycznego problemu aproksymacji funkcji. Zbadano wpływ liczby uruchamianych procesów na efektywność procedury uczenia i działania sieci oraz zademonstrowano negatywny wpływ opóźnień powstałych przy przesyłaniu danych za pomocą sieci LAN.
    In the paper some characteristic features concerning feed-forward neural network implementation in parallel computer architecture using MPI communication protocol are investigated. Two fundamental methods of neural network parallelization are described: neural (Fig. 1) as well as synaptic parallelization (Fig. 2). Based on the presented methods, an original application implementing feed-forward multilayer neural network was built. The application includes: a Java runtime interface (Fig. 3) and a computational module based on the MPI communication protocol. The simulation tests consisted in neural network application to classical problem of nonlinear function approximation. Effect of the number of processes on the network learning efficiency was examined (Fig. 4, Tab. 1). The negative effect of transmission time delays in the LAN is also demonstrated in the paper. The authors conclude that computational advantages of neural networks parallelization on a heterogeneous cluster consisting of several personal computers will become apparent only in the case of very complex neural networks, composed of many thousands of neurons.
    Źródło:
    Pomiary Automatyka Kontrola; 2011, R. 57, nr 6, 6; 638-640
    0032-4140
    Pojawia się w:
    Pomiary Automatyka Kontrola
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Estymacja błędów modelu powierzchni opisanych funkcjami kształtu za pomocą sieci neuronowych
    The estimation of errors of area models described by the shape functions by the means of neural networks
    Autorzy:
    Mrówczyńska, M.
    Powiązania:
    https://bibliotekanauki.pl/articles/341297.pdf
    Data publikacji:
    2007
    Wydawca:
    Uniwersytet Przyrodniczy we Wrocławiu
    Tematy:
    sieci neuronowe jednokierunkowe
    algorytmy gradientowe
    aproksymacja powierzchni
    neural network
    gradient methods of optimalization
    approximation method
    Opis:
    W artykule przedstawiono zagadnienie estymacji błędów modeli powierzchni określonej na dyskretnym zbiorze punktów o danych wartościach współrzędnych przestrzennych (x,y,z). Przyjęto, że obiekt opisują funkcje kształtu w postaci płaszczyzny, paraboloidy eliptycznej oraz paraboloidy hiperbolicznej. Realizacja numeryczna zadania polegała na wyznaczeniu błędów modeli określonych za pomocą sieci neuronowych oraz na podstawie rozwiązania zadań wyrównawczych. Modelowanie za pomocą sieci neuronowych zrealizowano za pomocą sieci jednokierunkowych wielowarstwowych z zastosowaniem gradientowych metod optymalizacji oraz algorytmu Resilientback Propagation (RPROP). Wyniki porównano z wynikami aproksymacji wielomianem drugiego i trzeciego stopnia, funkcją sklejaną oraz metodą kriging.
    The article deals with the issue of estimation of the area models errors determined on the basis of a discrete points set with the given values of space coordinates (x, y, z). The object was assumed to be described by shape functions in the form of the elliptic paraboloid and the hyperbolic paraboloid. The digital task accomplishment consisted in the statistic verification of errors of the models determined by neural networks and by the accomplishment of adjustment tasks. Modeling by the means of neural networks was carried out by the unidirectional multilayer networks with the application of gradient methods of optimalization and by Resilientback Propagation algorithm (RPROP). The obtained results were compared with the following results of approximation of the second and the third degree of polynomial, the b-spline function and the kriging's method.
    Źródło:
    Acta Scientiarum Polonorum. Geodesia et Descriptio Terrarum; 2007, 6, 1; 15-23
    1644-0668
    Pojawia się w:
    Acta Scientiarum Polonorum. Geodesia et Descriptio Terrarum
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    Applying ad hoc algorithms for highway traffic management
    Autorzy:
    Kabarowski, J.
    Powiązania:
    https://bibliotekanauki.pl/articles/393250.pdf
    Data publikacji:
    2010
    Wydawca:
    Polskie Stowarzyszenie Telematyki Transportu
    Tematy:
    telematyka transportu
    wybory lidera
    wielkość zbliżenia
    transport telematics
    leader election
    single hop network
    size approximation
    Opis:
    The number of cars participating in highway traffic is still growing as well as the necessity of quick and fluent transport between different locations. Therefore, ensuring the fluency and safety in highway traffic is becoming an essential problem in recent times. If only the driver knew about a danger waiting behind the bend, he would slow down; if someone got the information about a traffic jam 10 km earlier, he would probably take a different road or stop. This paper aims at providing transport telematics solutions which ensure simple, fast and efficient means for highway traffic management. The ad hoc algorithms presented concern broadcasting within highway traffic and assume the collisions detection and a single hop network. Installing a simple transmitter and receiver on every car is considered. Energy efficient size approximation algorithms and a leader selection procedure are presented and discussed, including simulations.
    Źródło:
    Archives of Transport System Telematics; 2010, 3, 2; 13-17
    1899-8208
    Pojawia się w:
    Archives of Transport System Telematics
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    An enhanced differential evolution algorithmwith adaptive weight bounds for efficient training ofneural networks
    Ulepszony algorytm ewolucji różnicowej z adaptacyjnymi granicami wag dla efektywnego szkolenia sieci neuronowych
    Autorzy:
    Limtrakul, Saithip
    Wetweerapong, Jeerayut
    Powiązania:
    https://bibliotekanauki.pl/articles/27315365.pdf
    Data publikacji:
    2023
    Wydawca:
    Politechnika Lubelska. Wydawnictwo Politechniki Lubelskiej
    Tematy:
    neural network
    differential evolution
    training neural network
    function approximation
    sieć neuronowa
    ewolucja różnicowa
    trening sieci neuronowej
    aproksymacja funkcji
    Opis:
    Artificial neural networks are essential intelligent tools for various learning tasks. Training them is challenging due to the nature of the data set, many training weights, and their dependency, which gives rise to a complicated high-dimensional error function for minimization. Thus, global optimization methods have become an alternative approach. Many variants of differential evolution (DE) have been applied as training methods to approximate the weights of a neural network. However, empirical studies show that they suffer from generally fixed weight bounds. In this research, we propose an enhanced differential evolution algorithm with adaptive weight bound adjustment (DEAW) for the efficient training of neural networks. The DEAW algorithm uses small initial weight bounds and adaptive adjustment in the mutation process. It gradually extends the bounds when a component of a mutant vector reaches its limits. We also experiment with using several scales of an activation function with the DEAW algorithm. Then, we apply the proposed method with its suitable setting to solve function approximation problems. DEAW can achieve satisfactory results compared to exact solutions.
    Sztuczne sieci neuronowe są niezbędnymi inteligentnymi narzędziami do realizacji różnych zadań uczenia się. Ich szkolenie stanowi wyzwanie ze względu na charakter zbioru danych, wiele wag treningowych i ich zależności, co powoduje powstanie skomplikowanej, wielowymiarowej funkcji błędu do minimalizacji. Dlatego alternatywnym podejściem stały się metody optymalizacji globalnej. Wiele wariantów ewolucji różnicowej (DE) zostało zastosowanych jako metody treningowe do aproksymacji wag sieci neuronowej. Jednak badania empiryczne pokazują, że cierpią one z powodu ogólnie ustalonych granic wag. W tym badaniu proponujemy ulepszony algorytm ewolucji różnicowej z adaptacyjnym dopasowaniem granic wag (DEAW) dla efektywnego szkolenia sieci neuronowych. Algorytm DEAW wykorzystuje małe początkowe granice wag i adaptacyjne dostosowanie w procesie mutacji. Stopniowo rozszerza on granice, gdy składowa wektora mutacji osiąga swoje granice. Eksperymentujemy również z wykorzystaniem kilku skal funkcji aktywacji z algorytmem DEAW. Następnie, stosujemy proponowaną metodę z jej odpowiednim ustawieniem do rozwiązywania problemów aproksymacji funkcji. DEAW może osiągnąć zadowalające rezultaty w porównaniu z rozwiązaniami dokładnymi.
    Źródło:
    Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska; 2023, 13, 1; 4--13
    2083-0157
    2391-6761
    Pojawia się w:
    Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    An artifical neural network usage in measurement of exhaust gas emission from marine engines: case study
    Autorzy:
    Kowalski, J.
    Rudzki, K.
    Powiązania:
    https://bibliotekanauki.pl/articles/2073611.pdf
    Data publikacji:
    2016
    Wydawca:
    Politechnika Gdańska. Wydział Inżynierii Mechanicznej i Okrętownictwa
    Tematy:
    marine engine
    emission
    neural network
    measurement
    power approximation
    silnik okrętowy
    emisja
    sieć neuronowa
    pomiar
    Opis:
    The paper presents the case study of use the artificial neural network to predict the main propulsion marine engine load. Mentioned load of the engine is important parameter to assess the emission level of toxic compounds into the atmosphere according to ISO standard and MARPOL convention. The engine load depends on the ship speed, rotational speed of the engine, propeller blades settings, the direction and the speed of wind, the condition of sea and the direction and the speed of sea currents and construction parameters of the ship. The realization of the aim of the work requires the direct measurement of presented parameters and measurement of exhaust gas composition. The experiment was carried out onboard STS “Pogoria”. Obtained results are enough to use the ANN to predict engine load to measure the emission level of toxic compounds.
    Źródło:
    Journal of Polish CIMEEAC; 2016, 11, 1; 87--93
    1231-3998
    Pojawia się w:
    Journal of Polish CIMEEAC
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    A novel fast feedforward neural networks training algorithm
    Autorzy:
    Bilski, Jarosław
    Kowalczyk, Bartosz
    Marjański, Andrzej
    Gandor, Michał
    Zurada, Jacek
    Powiązania:
    https://bibliotekanauki.pl/articles/2031099.pdf
    Data publikacji:
    2021
    Wydawca:
    Społeczna Akademia Nauk w Łodzi. Polskie Towarzystwo Sieci Neuronowych
    Tematy:
    neural network training algorithm
    QR decomposition
    Givens rotations
    approximation
    classification
    Opis:
    In this paper1 a new neural networks training algorithm is presented. The algorithm originates from the Recursive Least Squares (RLS) method commonly used in adaptive filtering. It uses the QR decomposition in conjunction with the Givens rotations for solving a normal equation - resulting from minimization of the loss function. An important parameter in neural networks is training time. Many commonly used algorithms require a big number of iterations in order to achieve a satisfactory outcome while other algorithms are effective only for small neural networks. The proposed solution is characterized by a very short convergence time compared to the well-known backpropagation method and its variants. The paper contains a complete mathematical derivation of the proposed algorithm. There are presented extensive simulation results using various benchmarks including function approximation, classification, encoder, and parity problems. Obtained results show the advantages of the featured algorithm which outperforms commonly used recent state-of-the-art neural networks training algorithms, including the Adam optimizer and the Nesterov’s accelerated gradient.
    Źródło:
    Journal of Artificial Intelligence and Soft Computing Research; 2021, 11, 4; 287-306
    2083-2567
    2449-6499
    Pojawia się w:
    Journal of Artificial Intelligence and Soft Computing Research
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
    Tytuł:
    A Deep Q-Learning Network for ship stowage planning problem
    Autorzy:
    Shen, Y.
    Zhao, N.
    Xia, M.
    Du, X.
    Powiązania:
    https://bibliotekanauki.pl/articles/260614.pdf
    Data publikacji:
    2017
    Wydawca:
    Politechnika Gdańska. Wydział Inżynierii Mechanicznej i Okrętownictwa
    Tematy:
    Deep Q-Leaning Network (DQN)
    container terminal
    ship stowage plan
    markov decision process
    value function approximation
    generalization
    Opis:
    Ship stowage plan is the management connection of quae crane scheduling and yard crane scheduling. The quality of ship stowage plan affects the productivity greatly. Previous studies mainly focuses on solving stowage planning problem with online searching algorithm, efficiency of which is significantly affected by case size. In this study, a Deep Q-Learning Network (DQN) is proposed to solve ship stowage planning problem. With DQN, massive calculation and training is done in pre-training stage, while in application stage stowage plan can be made in seconds. To formulate network input, decision factors are analyzed to compose feature vector of stowage plan. States subject to constraints, available action and reward function of Q-value are designed. With these information and design, an 8-layer DQN is formulated with an evaluation function of mean square error is composed to learn stowage planning. At the end of this study, several production cases are solved with proposed DQN to validate the effectiveness and generalization ability. Result shows a good availability of DQN to solve ship stowage planning problem.
    Źródło:
    Polish Maritime Research; 2017, S 3; 102-109
    1233-2585
    Pojawia się w:
    Polish Maritime Research
    Dostawca treści:
    Biblioteka Nauki
    Artykuł
      Wyświetlanie 1-15 z 15

      Ta witryna wykorzystuje pliki cookies do przechowywania informacji na Twoim komputerze. Pliki cookies stosujemy w celu świadczenia usług na najwyższym poziomie, w tym w sposób dostosowany do indywidualnych potrzeb. Korzystanie z witryny bez zmiany ustawień dotyczących cookies oznacza, że będą one zamieszczane w Twoim komputerze. W każdym momencie możesz dokonać zmiany ustawień dotyczących cookies