Rozprawy doktorskie na temat „Back propagation”
Kliknij ten link, aby zobaczyć inne rodzaje publikacji na ten temat: Back propagation.
Utwórz poprawne odniesienie w stylach APA, MLA, Chicago, Harvard i wielu innych
Sprawdź 50 najlepszych rozpraw doktorskich naukowych na temat „Back propagation”.
Przycisk „Dodaj do bibliografii” jest dostępny obok każdej pracy w bibliografii. Użyj go – a my automatycznie utworzymy odniesienie bibliograficzne do wybranej pracy w stylu cytowania, którego potrzebujesz: APA, MLA, Harvard, Chicago, Vancouver itp.
Możesz również pobrać pełny tekst publikacji naukowej w formacie „.pdf” i przeczytać adnotację do pracy online, jeśli odpowiednie parametry są dostępne w metadanych.
Przeglądaj rozprawy doktorskie z różnych dziedzin i twórz odpowiednie bibliografie.
1
Lowton, Andrew D. "A constructive learning algorithm based on back-propagation". Thesis, Aston University, 1995. http://publications.aston.ac.uk/10663/.
Pełny tekst źródłaStreszczenie:
There are been a resurgence of interest in the neural networks field in recent years, provoked in part by the discovery of the properties of multi-layer networks. This interest has in turn raised questions about the possibility of making neural network behaviour more adaptive by automating some of the processes involved. Prior to these particular questions, the process of determining the parameters and network architecture required to solve a given problem had been a time consuming activity. A number of researchers have attempted to address these issues by automating these processes, concentrating in particular on the dynamic selection of an appropriate network architecture. The work presented here specifically explores the area of automatic architecture selection; it focuses upon the design and implementation of a dynamic algorithm based on the Back-Propagation learning algorithm. The algorithm constructs a single hidden layer as the learning process proceeds using individual pattern error as the basis of unit insertion. This algorithm is applied to several problems of differing type and complexity and is found to produce near minimal architectures that are shown to have a high level of generalisation ability. (DX 187, 339)
2
Sanner, Robert M. (Robert Michael). "Neuromorphic regulation of dynamic systems using back propagation networks". Thesis, Massachusetts Institute of Technology, 1988. http://hdl.handle.net/1721.1/34995.
Pełny tekst źródła
3
Fernando, Thudugala Mudalige K. G. "Hydrological applications of MLP neural networks with back-propagation". Thesis, Hong Kong : University of Hong Kong, 2002. http://sunzi.lib.hku.hk/hkuto/record.jsp?B25085517.
Pełny tekst źródła
4
Bennett, Richard Campbell. "Classification of underwater signals using a back-propagation neural network". Monterey, Calif. : Springfield, Va. : Naval Postgraduate School ; Available from National Technical Information Service, 1997. http://handle.dtic.mil/100.2/ADA331774.
Pełny tekst źródłaStreszczenie:
Thesis (M.S. in Electrical Engineering) Naval Postgraduate School, June 1997.Thesis advisors, Monique P. Fargues, Roberto Cristi. Includes bibliographical references (p. 95). Also available online.
5
Ramachandran, Adithya. "HEV fuel optimization using interval back propagation based dynamic programming". Thesis, Georgia Institute of Technology, 2016. http://hdl.handle.net/1853/55054.
Pełny tekst źródłaStreszczenie:
In this thesis, the primary powertrain components of a power split hybrid electric vehicle are modeled. In particular, the dynamic model of the energy storage element (i.e., traction battery) is exactly linearized through an input transformation method to take advantage of the proposed optimal control algorithm. A lipschitz continuous and nondecreasing cost function is formulated in order to minimize the net amount of consumed fuel. The globally optimal solution is obtained using a dynamic programming routine that produces the optimal input based on the current state of charge and the future power demand. It is shown that the global optimal control solution can be expressed in closed form for a time invariant and convex incremental cost function utilizing the interval back propagation approach. The global optimality of both time varying and invariant solutions are rigorously proved. The optimal closed form solution is further shown to be applicable to the time varying case provided that the time variations of the incremental cost function are sufficiently small. The real time implementation of this algorithm in Simulink is discussed and a 32.84 % improvement in fuel economy is observed compared to existing rule based methods.
6
Teo, Chin Hock. "Back-propagation neural networks in adaptive control of unknown nonlinear systems". Thesis, Monterey, California. Naval Postgraduate School, 1991. http://hdl.handle.net/10945/26898.
Pełny tekst źródłaStreszczenie:
Approved for public release; distribution is unlimitedThe objective of this research is to develop a Back-propagation Neural Network (BNN) to control certain classes of unknown nonlinear systems and explore the network's capabilities. The structure of the Direct Model Reference Adaptive Controller (DMRAC) for Linear Time Invariant (LTI) systems with unknown parameters is first analyzed. This structure is then extended using a BNN for adaptive control of unknown nonlinear systems. The specific structure of the BNN DMRAC is developed for control of four general classes of nonlinear systems modeled in discrete time. Experiments are conducted by placing a representative system from each class under the BNN's control. The condition under which the BNN DMRAC can successfully control these systems are investigated. The design and training of the BNN are also studied. The results of the experiments show that the BNN DMRAC works for the representative systems considered, while the conventional least-squares estimator DMRAC fails. Based on analysis and experimental findings, some genera conditions required to ensure that this technique works are postulated and discussed. General guidelines used to achieve the stability of the BNN learning process and good learning convergence are also discussed. To establish this as a general and significant control technique, further research is required to obtain analytically, the conditions for stability of the controlled system, and to develop more specific rules and guidelines in the BNN design and training.
7
Cakarcan, Alpay. "Back-propagation neural networks in adaptive control of unknown nonlinear systems". Thesis, Monterey, California. Naval Postgraduate School, 1994. http://hdl.handle.net/10945/30830.
Pełny tekst źródłaStreszczenie:
The objective of this thesis research is to develop a Back-Propagation Neural Network (BNN) to control certain classes of unknown nonlinear systems and explore the network's capabilities. The structure of the Direct Model Reference Adaptive Controller (DMRAC) for Linear Time Invariant (LTI) systems with unknown parameters is first analyzed and then is extended to nonlinear systems by using BNN, Nonminimum phase systems, both linear and nonlinear, have also be considered. The analysis of the experiments shows that the BNN DMRAC gives satisfactory results for the representative nonlinear systems considered, while the conventional least-squares estimator DMRAC fails. Based on the analysis and experimental findings, some general conditions are shown to be required to ensure that this technique is satisfactory. These conditions are presented and discussed. It has been found that further research needs to be done for the nonminimum phase case in order to guarantee stability and tracking. Also, to establish this as a more general and significant control technique, further research is required to develop more specific rules and guidelines for the BNN design and training.
8
Xiao, Nancy Y. (Nancy Ying). "Using the modified back-propagation algorithm to perform automated downlink analysis". Thesis, Massachusetts Institute of Technology, 1996. http://hdl.handle.net/1721.1/40206.
Pełny tekst źródłaStreszczenie:
Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 1996.Includes bibliographical references (p. 121-122).
by Nancy Y. Xiao.
M.Eng.
9
Le, Chau Giang. "Application of a back-propagation neural network to isolated-word speech recognition". Thesis, Monterey, Calif. : Springfield, Va. : Naval Postgraduate School ; Available from National Technical Information Service, 1993. http://handle.dtic.mil/100.2/ADA272495.
Pełny tekst źródła
10
Chen, Peng. "Analysis of contribution rates and prediction based on back propagation neural networks". Thesis, University of Macau, 2017. http://umaclib3.umac.mo/record=b3691340.
Pełny tekst źródła
11
Guo, Xinxin. "Back-propagation beamformer design with transverse oscillations for motion estimation in echocardiography". Thesis, Lyon, INSA, 2014. http://www.theses.fr/2014ISAL0085/document.
Pełny tekst źródłaStreszczenie:
L'échographie est aujourd'hui l'une des modalités les plus populaires de diagnostic médical. Il permet d'observer, en temps réel, le mouvement des organes qui facilite le diagnostic des pathologies pour des médecins. L'échocardiographie [1, 2], l'imagerie du flux sanguin [3, 4] et l’élastographie [5-7] sont les domaines préférés de l'estimation de mouvement en utilisant l'échographie (en raison de son haut frame-rate).En conséquence, les images avec meilleurs qualités sont nécessaires. . En imagerie cardiaque, le système classique d'imagerie est limité dans la direction transversale (la direction perpendiculaire à celle de propagation). Travaillant sur la formation des images, ce problème peut être résolu en modifiant la façon de formateur de voie afin d'introduire des oscillations transversales (OTs) dans la fonction d’étalement du point (PSF). La technique d’oscillation transversale a montré son potentiel d'améliorer la précision de l'estimation de mouvement local dans la direction transversale (la direction perpendiculaire à celle de propagation). La classique OT en géométrie linéaire, basée sur l'approximation de Fraunhofer, relie la PSF et la fonction de pondération par la transformée de Fourier. Motivé par l'adaptation des OTs en échocardiographie, nous proposons une technique spécifique basée sur la rétro-propagation afin de construire des OTs en géométrie sectorielle. La performance de la méthode de rétro-propagation proposée a été étudiée progressivement, comparée avec la méthode de la transformée de Fourier, par exemple, l'évaluation de la qualité de la PSF quantifié, dans l'estimation de mouvement cardiaque en simulation, et en étude la qualité des PSF visuellement expérimentale. Les résultats quantifiés montrent les OT-images sont mieux contrôlés par la méthode proposée que par le formateur de voie conventionnelle. Une autre méthode, basée sur la décomposition d'onde plane et un principe différent de rétro-propagation, a été présentée. Cette méthode mieux prend en compte la propriété 2D de PSF, en décomposant la PSF dans un ensemble d'ondes planes directionnelle, les rétro-propage à la sonde, en utilisant les résultats de superposition comme excitations, un PSF simulée et conforme fortement au PSF théorique est acquis. En adaptant cette méthode à la géométrie sectorielle, la qualité de la PSF obtenue en face et sur la côté de la sonde est meilleure en utilisant la décomposition en ondes planes à celle de la transformée de Fourier, le travail supplémentaire sera adressé à adapter la décomposition en ondes planes à imagerie sectorielle et l’estimation du mouvementEchography is nowadays one of the most popular medical diagnosis modalities. It enables real-time observation the motion of moving organs which facilitates the diagnosis of pathologies for physician. Echocardiography [1, 2], blood flow imaging [3, 4] and elastography [5-7] are the favorite domains of motion estimation in using of echography (e.g., due to its high frame-rate capacity). Thus the requirements for imaging with high quality are on the primary place. In cardiac imaging, the conventional imaging system is somehow limited in the transverse direction (the direction perpendicular to the beam axis). Working on the image formation, this problem can be addressed by modifying the beamforming scheme in order to introduce transverse oscillations (TOs) in the system point spread function (PSF). Transverse oscillation techniques have shown their potential for improving the accuracy of local motion estimation in the transverse direction (i.e., the direction perpendicular to the beam axis). The conventional design of TOs in linear geometry, which is based on the Fraunhofer approximation, relates PSF and apodization function through a Fourier transform. Motivated by the adaptation of TOs in echocardiography, we propose a specific beamforming approach based on back-propagation in order to build TOs in sectorial geometry. The performance of the proposed back-propagation method has been studied gradually, in comparison with the Fourier transform, such as in evaluation of the quality of PSF, in estimation of simulated cardiac motion and in experiments study, etc. The quantified results demonstrate the proposed method leads to better controlled TOs images than the conventional beamforming. Another method based on plane wave decomposition and a different back-propagation principle has been presented. This method is better taking into account the 2D property of PSF, by decomposing the PSF into a set of plane waves directionally, back-propagating them to the probe, by using the superposition results as excitations, a simulated PSF with high accordance to the theoretical one is acquired. By adapting this method to sectorial geometry, the quality of PSF obtained in front of probe is better using the plane wave decomposition method than that of Fourier relation, but it is limited for the scanning on the side of probe, so the further work will be addressed to adapting the plane wave decomposition method to the complete sectorial imaging
12
Rose, Stephen Matthew. "Online training of a neural network controller by improved reinforcement back-propagation". Thesis, Georgia Institute of Technology, 2002. http://hdl.handle.net/1853/19177.
Pełny tekst źródła
13
WEI-LUN, CHENG, i 程韋綸. "Differential statistical method of Back-propagation neural networks and Grey-Box Back-propagation Network (GBPN)". Thesis, 2007. http://ndltd.ncl.edu.tw/handle/68211364989890317226.
Pełny tekst źródłaStreszczenie:
碩士中華大學
資訊管理學系
95
Although back-propagation neural networks (BPN) may construct accurate non-linear model, it belongs to black box model, and can not directly quantitatively measure the effect and importance of each input variable. To improve this shortcoming, this study proposed two methods: (1) Differential statistical method (DSM): We derived formula for the linear and quadratic differential of the input variable to the output variable on BPN, and after finished training network, calculated the differential of all training examples, and the average of them to be as the Linear Effect Index and Quadratic Effect Index. Besides, to quantitatively measure the importance of each input variable, General Important Index was defined as the root mean squared of the linear differential of all the training examples. (2) Grey-Box Back-propagation Network (GBPN): We modified the conventional BP algorithm, and Variable Importance Index was added on each input unit of input layer to quantitatively measure its importance to model, and the learning rule for the indexes was deduced with the gradient steepest descent method, which can adjust the indexes in the learning process, and then they can correctly measure the importance of input variable. By way of six artificial numerical examples and ten real application examples, it was demonstrated that (1) GBPN is slightly more accurate than BPN, and (2) these two methods can correctly measure the importance of input variables of linear, quadratic, and interactive functions, and DSM is more accurate than GBPN.
14
chu, liang-han, i 朱良瀚. "Back-Propagation Neural Network on Wave Forecast". Thesis, 1997. http://ndltd.ncl.edu.tw/handle/26611027074650855025.
Pełny tekst źródłaStreszczenie:
碩士國立中興大學
土木工程學系
85
In this paper, the back-propagation neural network (BPN) associated the I/O relationship in the Box-Jenkins model isestablished for the wave forecast model. The virtue of artificialneural network model is available for the short-term time series, thus it is useful for the wave prediction of offshore and coastal regions. A time series of Bretschneider wave spectrum performed in the laboratory is firstly adopted in this study to optimize the algorithm and network topology of the back-propagation neural network. The site wave data measured at Taichung Harbor and Kaoshiung LNG Port are then used to verify the accuracy of the model, based on the analysis of the efficiency coefficient, correlation coefficient and the root mean squared error between predicted and observed data. Waves of winter type and summer type are respectively simulated in the verification of model. The results show that the prediction has good performance in the winter waves when the short or longer training data is used. However, a longer training data should be utilized to have better performance for the summer waves due to storm waves being involved in the season.
15
Yang, Tsung-Lung, i 楊宗龍. "Engineering design optimization using back-propagation network". Thesis, 1994. http://ndltd.ncl.edu.tw/handle/41407894209919924542.
Pełny tekst źródła
16
Kuo, Hsiu-Min, i 郭秀敏. "Fuzzy Back Propagation in PCB Sale Forecasting". Thesis, 2004. http://ndltd.ncl.edu.tw/handle/81912708190658571530.
Pełny tekst źródłaStreszczenie:
碩士元智大學
工業工程與管理學系
92
Reliable prediction of sales can improve the quality of business strategy. In this research, fuzzy logic and artificial neural network are integrated into the fuzzy back-propagation network (FBPN) for printed circuit board industry. The fuzzy back propagation network is constructed to incorporate production-control expert judgments in enhancing the performance. Parameters chosen as inputs to the FBPN are no longer considered as of equal importance, but some production control experts are requested to express their opinions about the importance of each input parameter in predicting the sales with linguistic terms, which can be converted into pre-specified fuzzy numbers, aggregated and corresponding input parameter when fed into the FBPN. Subsequently, the arithmetic for triangular fuzzy numbers is applied to deal with all calculation involved in network learning. The proposed system is evaluated through the real life date provide by GCE company. Model evaluation results for research indicate that the Fuzzy back-propagation outperforms the other three different forecasting models in both MAPE and MAR.
17
Chen, Yun-Chien, i 陳芸仟. "Earthquake Prediction Via Back Propagation Neural Network". Thesis, 2013. http://ndltd.ncl.edu.tw/handle/87238218545102882108.
Pełny tekst źródłaStreszczenie:
碩士國立臺北教育大學
資訊科學系碩士班
101
Taiwan is located between Eurasian and Pacific plates, the seismicity is very frequently and actively. Of course it’s including more than 6.0 of the magnitude of the earthquake. Beside, the population rate is very high in Taiwan, so the disaster caused by the earthquake always loss human life and economic. Therefore the research of the earthquake precursor and earthquake prediction is an important issue. It’s the common and easier method that observes the change of the short time seismicity rate base on long time seismicity rate, like value, value, Z value, value, the cumulative magnitude and the quantity of the earthquakes. These methods can depict the seismicity well, but can not completely predict the arrival of a main earthquake. Scientists still can’t understand the mechanism that the earthquake occurs totally, but they believe that these parameters of the methods should be the nonlinear correlation. However, just we can unable to establish a good physical model to describe the occurrence mechanism of earthquakes. Back-propagation neural network, mimic the biological born neurons, that has the good performance for solving nonlinear problems without the prior predictable model. In this study, we attempt to put the value, value, Z value and quality of earthquakes number into the input layer of the back-propagation neural network. Then predict the largest magnitude in the next month. After train and test data from 1994 - 2011, forecast the largest capacity in the next month: 72% is success if the magnitude is between 5.0-6.5, 39% is success if the magnitude is more than 6.5.
18
Li, Jian-Ming, i 李健銘. "VLSI Layout of a Back-Propagation Neuro-Microprocessor". Thesis, 2006. http://ndltd.ncl.edu.tw/handle/68790467743608967730.
Pełny tekst źródłaStreszczenie:
碩士大葉大學
電機工程學系碩士班
94
This study develops a 32-bit RISC processor core embedded with first-order back-propagation neural network and MIPS-like architecture by using Verilog HDL and algorithmic state machine (ASM). The designed processor core is carried out through the behavioral stage by simulation of SynaptiCAD and synthesis of Xilinx FPGA development software. The VLSI layout of a neuro-microprocessor core is implemented under TSMC 0.18 um process technology at final.
19
"Dynamic construction of back-propagation artificial neural networks". Chinese University of Hong Kong, 1991. http://library.cuhk.edu.hk/record=b5886957.
Pełny tekst źródłaStreszczenie:
by Korris Fu-lai Chung.Thesis (M.Phil.) -- Chinese University of Hong Kong, 1991.
Bibliography: leaves R-1 - R-5.
LIST OF FIGURES --- p.vi
LIST OF TABLES --- p.viii
Chapter 1 --- INTRODUCTION
Chapter 1.1 --- Recent Resurgence of Artificial Neural Networks --- p.1-1
Chapter 1.2 --- A Design Problem in Applying Back-Propagation Networks --- p.1-4
Chapter 1.3 --- Related Works --- p.1-6
Chapter 1.4 --- Objective of the Research --- p.1-8
Chapter 1.5 --- Thesis Organization --- p.1-9
Chapter 2 --- MULTILAYER FEEDFORWARD NETWORKS (MFNs) AND BACK-PRO- PAGATION (BP) LEARNING ALGORITHM
Chapter 2.1 --- Introduction --- p.2-1
Chapter 2.2 --- From Perceptrons to MFNs --- p.2-2
Chapter 2.3 --- From Delta Rule to BP Algorithm --- p.2-6
Chapter 2.4 --- A Variant of BP Algorithm --- p.2-12
Chapter 3 --- INTERPRETATIONS AND PROPERTIES OF BP NETWORKS
Chapter 3.1 --- Introduction --- p.3-1
Chapter 3.2 --- A Pattern Classification View on BP Networks --- p.3-2
Chapter 3.2.1 --- Pattern Space Interpretation of BP Networks --- p.3-2
Chapter 3.2.2 --- Weight Space Interpretation of BP Networks --- p.3-3
Chapter 3.3 --- Local Minimum --- p.3-5
Chapter 3.4 --- Generalization --- p.3-6
Chapter 4 --- GROWTH OF BP NETWORKS
Chapter 4.1 --- Introduction --- p.4-1
Chapter 4.2 --- Problem Formulation --- p.4-1
Chapter 4.3 --- Learning an Additional Pattern --- p.4-2
Chapter 4.4 --- A Progressive Training Algorithm --- p.4-4
Chapter 4.5 --- Experimental Results and Performance Analysis --- p.4-7
Chapter 4.6 --- Concluding Remarks --- p.4-16
Chapter 5 --- PRUNING OF BP NETWORKS
Chapter 5.1 --- Introduction --- p.5-1
Chapter 5.2 --- Characteristics of Hidden Nodes in Oversized Networks --- p.5-2
Chapter 5.2.1 --- Observations from an Empirical Study --- p.5-2
Chapter 5.2.2 --- Four Categories of Excessive Nodes --- p.5-3
Chapter 5.2.3 --- Why are they excessive ？ --- p.5-6
Chapter 5.3 --- Pruning of Excessive Nodes --- p.5-9
Chapter 5.4 --- Experimental Results and Performance Analysis --- p.5-13
Chapter 5.5 --- Concluding Remarks --- p.5-19
Chapter 6 --- DYNAMIC CONSTRUCTION OF BP NETWORKS
Chapter 6.1 --- A Hybrid Approach --- p.6-1
Chapter 6.2 --- Experimental Results and Performance Analysis --- p.6-2
Chapter 6.3 --- Concluding Remarks --- p.6-7
Chapter 7 --- CONCLUSIONS --- p.7-1
Chapter 7.1 --- Contributions --- p.7-1
Chapter 7.2 --- Limitations and Suggestions for Further Research --- p.7-2
REFERENCES --- p.R-l
APPENDIX
Chapter A.1 --- A Handwriting Numeral Recognition Experiment: Feature Extraction Technique and Sampling Process --- p.A-1
Chapter A.2 --- Determining the distance d= δ2/2r in Lemma 1 --- p.A-2
20
"A multiresolution learning method for back-propagation networks". Chinese University of Hong Kong, 1994. http://library.cuhk.edu.hk/record=b5887215.
Pełny tekst źródłaStreszczenie:
Wing-Chung Chan.Thesis (M.Phil.)--Chinese University of Hong Kong, 1994.
Includes bibliographical references (leaves 81-85).
Chapter 1 --- Introduction --- p.1
Chapter 2 --- Multiresolution Signal Decomposition --- p.5
Chapter 2.1 --- Introduction --- p.5
Chapter 2.2 --- Laplacian Pyramid --- p.6
Chapter 2.2.1 --- Gaussian Pyramid Generation --- p.7
Chapter 2.2.2 --- Laplacian Pyramid Generation --- p.7
Chapter 2.2.3 --- Decoding --- p.8
Chapter 2.2.4 --- Limitation --- p.9
Chapter 2.3 --- Multiresolution Transform --- p.9
Chapter 2.3.1 --- Multiresolution Approximation of L2(R) --- p.9
Chapter 2.3.2 --- Implementation of a Multiresolution Transform --- p.12
Chapter 2.3.3 --- Orthogonal Wavelet Representation --- p.16
Chapter 2.3.4 --- Implementation of an Orthogonal Wavelet Representation --- p.18
Chapter 2.3.5 --- Signal Reconstruction --- p.21
Chapter 3 --- Multiresolution Learning Method --- p.23
Chapter 3.1 --- Introduction --- p.23
Chapter 3.2 --- Input Vector Representation --- p.24
Chapter 3.2.1 --- Representation at the resolution 1 --- p.24
Chapter 3.2.2 --- Representation at the resolution 2j --- p.25
Chapter 3.2.3 --- Border Problem --- p.26
Chapter 3.3 --- Back-Propagation Network Architecture --- p.26
Chapter 3.4 --- Training Procedure Strategy --- p.27
Chapter 3.4.1 --- Sum Squared Error (SSE) --- p.28
Chapter 3.4.2 --- Intermediate Stopping Criteria --- p.30
Chapter 3.5 --- Connection Weight Transformation --- p.31
Chapter 3.5.1 --- Weights between the Input and Hidden Layers --- p.31
Chapter 3.5.2 --- Weights between the Hidden and Output Layers --- p.33
Chapter 4 --- Simulations --- p.36
Chapter 4.1 --- Introduction --- p.36
Chapter 4.2 --- Choices of the Impulse Response h(n) --- p.36
Chapter 4.3 --- XOR Problem --- p.39
Chapter 4.3.1 --- Setting of Experiments --- p.39
Chapter 4.3.2 --- Experimental Results --- p.41
Chapter 4.4 --- Numeric Recognition Problem --- p.50
Chapter 4.4.1 --- Setting of Experiments --- p.50
Chapter 4.4.2 --- Experimental Results --- p.52
Chapter 4.5 --- Discussions --- p.72
Chapter 5 --- Conclusions --- p.75
Chapter A --- Proof of Equation (4.9) --- p.77
Chapter B --- Proof of Equation (4.11) --- p.79
Bibliography --- p.81
21
Jeng, Jia-Haur, i 鄭家豪. "Improved back-propagation networks for reservoir inflow forecasting". Thesis, 2008. http://ndltd.ncl.edu.tw/handle/81929178641702467805.
Pełny tekst źródłaStreszczenie:
碩士國立臺灣大學
土木工程學研究所
96
The efficiency is an important issue for neural networks-based models, but the issue has received little attention in the hydrologic domain. Back-propagation networks (BPNs) are the most frequently used convectional neural networks (NNs). However, BPNs are trained by the error back-propagation algorithm which is a very time-consuming iterative process. To improve the efficiency, improved BPNs which are trained by a novel query learning approach are proposed. The proposed query learning approach is capable of selecting informative data from all training data. Then the improve BPNs can be efficiently trained with partial data. An application is conducted to demonstrate the superiority of the improved BPNs. Two kinds of BPN-based (the improved and the conventional BPN-based) reservoir inflow forecasting models are constructed and the comparison between the improved and the conventional BPN-based model is made. The results show that the performance of the improved BPN-based models is as good as that of the conventional BPN-based models, but the improved BPN-based models significantly required less training time than the conventional BPN-based models. As compared to the conventional BPN models, only about 50% of training time is required for the improved BPN-based models. The improved BPN-based models are recommended as an alternative to the existing models because of their efficiency.
22
Wang, Nan-Ching, i 王南景. "Dynamic Back-Propagation for Plant Identification and Control". Thesis, 1994. http://ndltd.ncl.edu.tw/handle/09703957867238072262.
Pełny tekst źródłaStreszczenie:
碩士國立交通大學
控制工程系
82
While much of the recent emphasis in the connectionist reseaarch has been on feedforward networks with static back- propagation, it is likely that the use of dynamic networks will be of particular importance in control-related applications. This thesis is focused on a learning methodology for recurrent networks with feedback connections and feedforward networks as subsystems in a dynamic system. Such a learning methodology is termed dynamic back-propagation, which is one of the most prominent learning methods for connectionist networks. A detailed study of dynamic back-propagation is presented to provide an insight of the principal ideas that contributed to the evolution of the concept and the details concerning its practical applications to identification and control.
23
Chen, Chang-Chieh, i 陳昌捷. "Stock Index Prediction Using Back Propagation Neural Networks". Thesis, 2015. http://ndltd.ncl.edu.tw/handle/47651467208536124110.
Pełny tekst źródłaStreszczenie:
碩士國立宜蘭大學
多媒體網路通訊數位學習碩士在職專班
103
Abstract The main purpose of this study is to construct a Back Propagation Neural Network ( BPNN) model on MATLAB for predicting the Taiwan stock exchange capitalization weighted stock (TAIEX). The data ranging from 2014.01.02 to 2014.07.31 is selected. The duration is 7 months and there are total 140 recorders. The weighted indexes and technical analysis indicators are screened as the input parameters by using Pearson correlation coefficient. Specifically, the indicators, that the r values are more than 0.7, are selected as the input parameters, and there are total 17 input parameters. The input parameters are divided into three groups where the r values are 0.7, 0.8, and 0.9, respectively. Finally, Mean Absolute Percentage Error (MAPE) is used to evaluate the accuracy of the models. The results show that the MAPE of the prediction for index closed point is 0.6315% . In the short term (8 days) prediction, the accuracy is up to 87.5%. The accuracy of the short term prediction trends for the weighted index is 71.42%。
24
Chen, Wei-Yu, i 陳韋佑. "Development and Application ofDecision Group Back-Propagation Network". Thesis, 2003. http://ndltd.ncl.edu.tw/handle/8k7c43.
Pełny tekst źródłaStreszczenie:
碩士逢甲大學
土木及水利工程所
91
To apply deterministic model for system simulation, the assumption that systematic input-output relationship should be held in the whole event is often required. However, this assumption does not apply for nonlinear time-variant system which possible different outputs can be found for the various input structures. Under such circumstance, decision maker has to face the risk of system uncertainty which a deterministic model falls to handle in forecasting application. For this reason, this study first utilized Back-Propagation Network (BPN) as main structure to develop Decision Group Back-Propagation Network (DGBPN) in order to create numerous BPN models which are qualified by the accuracy criterion of fitting learning data. The model then chooses the suitable model(s) from these BPN models to compute their corresponding outputs such that this model can avoid the risk of forecasting by single deterministic model. With validation tests at Wu-Xi watershed, DGBPN performed stably and concluded fair forecast results. Allover, this research developed the methodology to provide the decision maker with trustworthy on flood forecasting.
25
Lin, Jyh-Woei, i 林志偉. "Applying Back Propagation Neural Network to Earthquake Predication". Thesis, 2019. http://ndltd.ncl.edu.tw/handle/bnh65y.
Pełny tekst źródłaStreszczenie:
博士南臺科技大學
電機工程系
107
Three back propagation neural network (BPNN) methods were applied to meet the three objectives in this dissertation. An embedded earthquake Richter magnitude (ML) prediction BPNN (EEMPBPNN) was built to predict more accurate Richter magnitude (ML) of five earthquakes in 2016 and 2018 (Taiwan Standard Time, TST), which exhibited high accuracy due to the lower standard deviation (SDV), lower mean squared error (MSE) and higher correlation coefficient. The number of neurons in each hidden layer of the EEMPBPNN was determined using earthquake catalogues and a slip rate of 7–8 cm/y between the Philippine Sea plate and Eurasian plate as training data, where the number of neurons has not been determined by training data in previous studys. An objective of this dissertation is to use a new Elementary Modified Levenberg–Marquardt Algorithm (M-LMA) as the error back propagation (EBP) algorithm to minimise backpropagation errors in training a BPNN, which was predicted using four seismic records for Earthquake Early Warning (EEW) associated with the Chi-Chi earthquake. The real four seismic records from four seismic stations belonged to Free Field Strong Earthquake Observation Network of Central Weather Bureau (CWB). The predicted seismic records were compared with real seismic records. For the predicted errors, a Trade-Off Decision-Making Process with BPNN (TDPB) was used to adjust the threshold of the error amplitudes to increase the warning time of EEW. This approach was not necessary to consider the problems of characterising the wave phases, data pre-processing, and complex hardware of previous studies. An active probability BPNN model (PBNNM) was built to predict the probability distribution belonged to the probabilistic seismic hazard analysis (PSHA). The determining of the framework of PBNNM was similar to EEMPBPNN. The studied region was divided into 500 small grids, each 0.2° X 0.2° in size. Each grid was assigned a predicted earthquake occurrence probability by the PBNNM for a month between 2015 to 2018. A quantitative analysis of the predicted reliability of the PBNNM, the standard error of the mean (SEM), and the normalised mean square error (NMSE) were used to evaluate predicted probability errors of the PBNNM. These low SEM and NMSE values confirmed the high accuracy of the PBNNM. The PBNNM could be commercialised with relatively low cost compared with the methods used in previous studies. Keywords: Embedded Earthquake Richter magnitude (ML) Prediction BPNN (EEMPBPNN), Elementary Modified Levenberg–Marquardt Algorithm (M-LMA), Active Probability BPNN Model (PBNNM).
26
Bor-Shing, Lin. "Using Back-Propagation Neural Network for Automatic Wheezing Detection". 2005. http://www.cetd.com.tw/ec/thesisdetail.aspx?etdun=U0001-0707200615372500.
Pełny tekst źródła
27
Chan, Shih Wei, i 詹世偉. "Detection of Oil Reservoir Bright Spot Using Back-propagation". Thesis, 2008. http://ndltd.ncl.edu.tw/handle/838vz7.
Pełny tekst źródłaStreszczenie:
碩士國立交通大學
土木工程系所
96
he complexity of two-dimensional seismic data often leads to mistakes in discriminating oil reservoir. However, mis-drilling caused by these mistakes brings about thirty million loss in cost each time. Seismic data interpreters do recognition by rules and experiences. If we can use neural network in place of rules and experiences, then we can get rid of some chances of seismic data interpreters’ emotional discrimination or mislook. The past oil reservoir bright spot detections still can't be applied to practice for two reasons. First, target seismic data in past research to be detected is in perfect condition. Interpreters seldom deal with these kinds of seismic data in reality. There is a big difference between seismic data in perfect condition and seismic data in practice. Second, the seismic attributes that past research used differs from that interpreters used in oil companies. The primary goal of this research is to apply back-propagation neural network to pattern recognition of oil reservoir bright spot. By interviewing with Seismic data interpreters in the oil company, we propose five seismic attributes in common use including seismic signal, evelope, instantaneous frequency, instantaneous phase and inversion impedance. After five seismic attributes of feature extraction, we do pre-processing on extracted features including transformating .segy file into .mat file, elimination of blunder, elimination of outlier, normalization and building feature set matrixes. Then, we import feature set into neural network and train matrix by matrix. By tuning any possible neural network layer, hidden layer node, training function(Levenberg-Marquardt, Conjugate Gradient), we have summed up to 12480 times of neural network training. Finally, we propose a method of optimized-recognition rated oil reservoir bright spot detection.
28
Lin, Bor-Shing, i 林伯星. "Using Back-Propagation Neural Network for Automatic Wheezing Detection". Thesis, 2006. http://ndltd.ncl.edu.tw/handle/61724033247949569395.
Pełny tekst źródłaStreszczenie:
博士國立臺灣大學
電機工程學研究所
94
Wheezes are continuous adventitious lung sounds, which have been defined as lasting for at least 250 ms. They are probably produced when airflow makes narrow airways vibrate. Wheezes are common clinical signs in patients with obstructive pulmonary disease such as asthma. Automatic wheezing detection offers an objective and accurate way to analyze wheezing lung sounds. By the automatic wheezing detection system, the features about amount, time duration, frequency range, and power strength of wheezes can be extracted to help physicians diagnose. It also can provide long-term auscultation and analysis of a patient. This Dissertation describes the design of a fast and high performance wheeze recognition system. First, respiratory sounds are captured, amplified and filtered by an analog circuit; then digitized through a PC soundcard, and recorded in accordance with the Computerized Respiratory Sound Analysis (CORSA) standards. Since the proposed wheezing detection algorithm is based on three methods: 2D bilateral filtering of spectrogram, order truncate average (OTA) method, and moving average (MA) method. Some features are then extracted from the processed spectra to train a back-propagation neural network (BPNN). Eventually, the new testing samples go through the trained BPNN to recognize whether they are wheezing sounds. Experiment results of the MA method show a high sensitivity of 1.0 and a specificity of 0.895 in qualitative analysis of wheeze recognition. Due to its high efficiency, great performance and easy-to-implement features, this wheeze recognition system could be of interest in the clinical monitoring of asthma patients and the study of physiological mechanisms in the respiratory airways.
29
陳俊呈. "The Application of Back-Propagation Network in Financial Distress". Thesis, 1999. http://ndltd.ncl.edu.tw/handle/67910599159207336885.
Pełny tekst źródłaStreszczenie:
碩士國立海洋大學
航運管理學系
87
In the past researches of financial distress prediction, traditional statistical techniques such as multivariate statistical was the majority method. Before using the multivariable statistical method, we take factor analysis to find representative ratio of independent variables. But that might ignore some influential variables to the model. Accordingly, I decide use all variables as input variable to build models and preserve fullness information of it. There were been more and more artificial neural network applications to this field in domestic since 1994. According to those researches, financial distress prediction models build by artificial neural network was more feasible than traditional statistical methods. This paper applied back-propagation Network to build the financial distress prediction models and collects 9 different industries, 37 companies’ 23 financial ratio as input variables. The results indicate that the models build by back-propagation network can accurate predict financial distress companies.
30
Tsong, Jeng Jiin, i 鄭錦聰. "A study of a modified Back-Propagation Neural network". Thesis, 1993. http://ndltd.ncl.edu.tw/handle/31875992630384494145.
Pełny tekst źródłaStreszczenie:
碩士國立臺灣科技大學
工程技術研究所
81
To overcome the shortcomings of the Back-propagation (Bp) algorithm, namely, slow convergence, local minimum, and paralysis problems, a combined Back-propagation/Cauchy ems, a combined Back-propagation/ (Bp/Cauchy) machine has been proposed by Wasserman [33]. In this thesis, two topics are explored. one is the switching condition to improve the Back- propagation/Cauchy machines network, and the other is to determine the optimal number of hidden units.Firstly, a switching condition is introduced to improve the Back- propagation/Cauchy machines network. If the switching condition is not satisfied, then we use Bp/Cauchy learning algorithm. If the switching condition holds, then we switch back to use Bp learning algorithm. It is shown that, under the switching condition, the error function decreases asympototically. As a result, when the switching condition holds, the Bp algorithm alone will converges very quickly to its equilibrium points. Secondly,we propose an improved Bp/ Cauchy network to determine the optimal number of hidden layer neurons (Note that this structure is one hidden layer neural network.).The results indicate that the optimal number of hidden layer neurons depondent on the input patterns. If the input patterns are totally independent, then the optimal number of hidden layer neurons #p is equal to the number of input patterns P. To illustrate the effectiveness of the propose method, examples of xor, parity check of seven bits, learning of unknown function, and application to magnetic bearing systems, are included. Results show that the improved Bp/Cauchy machine.118
31
Lee, Ching-Yi, i 李靜宜. "Predicting the Wafer Yield Using Back-Propagation Neural Network". Thesis, 2003. http://ndltd.ncl.edu.tw/handle/15999099231906740934.
Pełny tekst źródłaStreszczenie:
碩士國立交通大學
工業工程與管理系
91
For integrated circuit (IC) manufacturers, the wafer yield is a key index to evaluate their profit. There are two major factors affecting the wafer yield. One factor is the number of defects on a wafer and the other factor is the degree of defect clustering. As the wafer size increases, the clustering phenomenon of the defects becomes increasingly apparent. In this case, the conventional Poisson yield model will frequently underestimate the actual wafer yield. Although many modified yield models have been developed, these models are too complicated for practical use. In this study, Back-Propagation Neural Network is employed to amend the conventional Poisson Yield model with a clustering index. A case provided by Taiwan IC Company is also presented to demonstrate the effectiveness of the proposed approach. Comparisons are also made among the conventional Poisson yield model, modified yield models and the proposed yield model.
32
Lai, Liang-Bin, i 賴良賓. "Data Mining by Query-Based Back-Propagation Neural Networks". Thesis, 2003. http://ndltd.ncl.edu.tw/handle/65290993455760105933.
Pełny tekst źródłaStreszczenie:
碩士國立中央大學
資訊管理研究所
91
The central focus of data mining in enterprises is to gain insight into large collections of data for making a good prediction and a right decision. Neural networks have been applied to a wide variety of problem domains such as steering motor vehicles, recognizing genes in uncharacterized DNA sequences, scheduling payloads for the space shuttle, and predicting exchange rates. Advantages of neural networks include the high tolerance to noisy data as well as the ability to classify patterns having not been trained. Neural networks have been successfully applied to a wide range of supervised and unsupervised learning problems. However, while being applied in data mining, there are two fundamental considerations - the comprehensibility of learned models and the time required to induce models from large data sets. For the first problem, many approaches have been proposed for extracting rules from trained neural networks. In this thesis, we focus on the second problem. We introduce a query-based learning algorithm to improve neural networks'' performance in data mining. Results show that the proposed algorithm can significantly reduce the training set cardinality. Our future work is to apply this learning procedure to other data mining schemes.
33
LIN, TZU-YANG, i 林子揚. "Back-Propagation Neural Networks System for Recogniting Japanese Character". Thesis, 2016. http://ndltd.ncl.edu.tw/handle/51868533944161090993.
Pełny tekst źródłaStreszczenie:
碩士國立高雄海洋科技大學
電訊工程研究所
104
Thanks to the advancement of technology, neural networks can be used widely and in different aspects. They are found everywhere in life but are most commonly used to predict, identify, classify and self-learn. In this study, backpropagation neural network is used for word identification in Japanese more specifically identifying Japanese fonts shown in different ways correctly. Nowadays, it is a globalized society, and language plays an important role in our life. Unfortunately, not every one of us is familiar to the languages of each country and thus we rely on metaphorical bridges which are called translators to communicate with people from different countries. However, as for the aspects related to business and communication, paperwork is always needed. Since everyone’s handwriting is distinctive, the main point discussed in the study is then how to identify certain unclear fonts. A backpropagation network is used to do symbol identification in this study. It is designed to identify Japanese Katakana by getting imaging systems to digitize each Japanese sound and collect them. They are categorized by visual range of the system using a five by seven boolean lattice to represent each sound. However, imaging systems are not complete enough and each sound may be interfered by the noise, so the program is designed to perfectly classify the ideal input vectors and reasonably and correctly classify the noise vectors. Keywords: Backpropagation Neural Network, MATLAB, font identification
34
Liu, Tse-Han, i 劉澤翰. "Implementation of FPGA-Based Back-Propagation Artificial Neural Network". Thesis, 2011. http://ndltd.ncl.edu.tw/handle/67071260972142504766.
Pełny tekst źródłaStreszczenie:
碩士國立交通大學
電控工程研究所
99
This thesis presents a hardware design of artificial neural network for learning on the Altera DE2-70 FPGA board with 50 MHz operation frequency. There are two modes developed for users, application and training, to execute the calculation of neural network and to learn the weights. For the application mode, the multilayer architecture is realized by the layer multiplexing, reusing a single layer. Besides, the activation function, log-sigmoid, is approximated by the PWL method, to reduce the resource and speed up the operation. As for the learning mode, the off-line training for back-propagation algorithm is adopted to adjust the weights. Based on sequential architecture, the design complexity and resource requirement is further reduced. The data format adopts 24-bits fixed-point and the structure of ANN could be reconfigured by the parameters concerning the number of neuron or hidden-layer. The success of the hardware architecture is demonstrated by the experiment results of neural network applied to M-G curve prediction and image edge detection.
35
Kuo, Gang-Yaw, i 郭功耀. "VLSI Design of Back Propagation Networks with On-Chip Learning". Thesis, 2002. http://ndltd.ncl.edu.tw/handle/88281184606281919119.
Pełny tekst źródłaStreszczenie:
碩士國立交通大學
電機與控制工程系
90
Nowadays, the industry of information appliances and communication products is growing rapidly. Intelligent products will become the key feature in the future. Artificial neural networks have the capabilities to learn and recall and are highly parallel. However, conventional computers do not support parallel computing and learning capability that are inherent in neural networks. Among the existing parallel architectures, SIMD (Single Instruction stream Multiple Data) is the most suitable for the implementation of BPN (back propagation networks). Therefore, the proposed architecture is based on SIMD. The proposed architecture uses limited number of PEs to fulfill all the operations needed for the recalling phase and the learning phase. The aim of the proposed architecture is not intended for one specific application. Therefore, the proposed BPN chip can be reconfigured to any BPN structure by modifying some parameters. Finally, two real cases are used to verify our design.
36
Yu, Chih-Yao, i 游智堯. "The Back Propagation Neural Network for Studying Taiwan Electronic Option". Thesis, 2006. http://ndltd.ncl.edu.tw/handle/95741985552697523308.
Pełny tekst źródłaStreszczenie:
碩士嶺東科技大學
財務金融研究所
94
One kind of information processing technology is neural networks which have been developed rapidly in recent years. Especially in the financial field, neural networks are very popular. Therefore, focused on the Back Propagation Neural (BPN) Network, the study is to predict the price of electronic option. Moreover, comparing with the theoretical price of the Black-Scholes (B-S) model, we found the most suitable model of forecasting to investors as the investment materials. By the empirical results, the Back Propagation Neural (BPN) Network is more capable of forecasting than the Black-Scholes (B-S) model either call or put. Besides, in the Back Propagation Neural (BPN) Network model, there is very great influence on the results through choosing the input variables. Moreover, the result of considering open interest variables comes more accurately than that of not considering open interest variables. By observing MAE, MSE or RMSE, we could get the forecast value relatively closed to the market price under the consideration of joining the open interest variables. However, if only choosing the relevant variables based on the Black-Scholes (B-S) model, we might not get better forecast value. That is, in the factors of influencing option price, besides stock price, exercise price, non-risk interest rate, volatility and the expiration date, the open interest variables have been an important reference index. Thus, the empirical results show that in the model with call and put, at money option price is more accurate than in-the-money and out-of-the-money ones. Unlike in-the-money and out-of-the-money options, we don’t consider the time value in the model with at money option. Therefore, there are not many errors appeared to forecast the option price in the model with at money option.
37
Shen, Jia-N., i 沈家恩. "Wave Forecasting and Supplementary using Back-Propagation Neural Network Model". Thesis, 1998. http://ndltd.ncl.edu.tw/handle/24758958069795237158.
Pełny tekst źródłaStreszczenie:
碩士國立中興大學
土木工程學系
86
Accurate prediction for the wave climate is an essential part in the ocean engineering. This paper reports an application of the artificial neural network for accurate forecast of waves from a time series of the field data. This paper also presents the wave supplementary for a wave record using the neural network model. The back-propagation procedures for minimizing the error of the desired output is used in the learning process of the neural network. The field data measured in three wave stations at Taichung Harbor were used to test theperformance of the neural network model. The wave prediction from the time series of one wave stationor two wave stations is presented. It is found that the neural network model performs well for the waveforecast and wave supplementary when a very short-term wave data is used as a training set. In general, the wave prediction or wave supplementaryfrom the time series of two wave stations has betterperformance than one wave station records used. It is also found that the performance of theprediction of significant wave heights isbetter than that of wave periods.
38
Pan, Jim, i 潘曉駿. "Application of the Taiwan stock market by back-propagation network". Thesis, 1995. http://ndltd.ncl.edu.tw/handle/16687926942973902799.
Pełny tekst źródła
39
Wei-Han, Chen, i 陳維翰. "Application of Back-Propagation Neural Network for BGA Inspection System". Thesis, 2004. http://ndltd.ncl.edu.tw/handle/64028445868082588821.
Pełny tekst źródłaStreszczenie:
碩士國立臺灣科技大學
自動化及控制研究所
92
The main purpose of this research is to apply back-propagation neural network and computer vision to develop a two-dimension BGA (Ball Grid Array) defect inspection system. By using this system, the automatic inspection via computer vision can reduce the human error. The developed inspection functions include ball position offset, ball size and ball shape. In this research, the information of solder balls can be obtained through the following steps: the image grabbing of solder balls, median filtering, the binary image by using the Otsu’s method, morphology image processing, blob analysis, subpixel edge-detect, and the best fit ellipse equation. The acquired information of each solder ball, such as center, area and axis length were processed into center offset, area ratio and axis ratio of ellipse for neural network use. Those data were used to train the back-propagation neural network. The coordinate processing was also considered to overcome the problems while the BGA with random shift or rotation. After the back-propagation neural network has been trained successfully, it can be used to inspect the solder balls to be examined. According to the self-learning and highly recognized capability of the BPNN, this BGA inspection system is accurate for inspecting the defect of solder balls. The system still can inspect correctly even though the BGA with shift or angle of rotation.
40
李其縵. "A research of applying back-propagation network to knowledge extraction". Thesis, 2003. http://ndltd.ncl.edu.tw/handle/45770850335995372926.
Pełny tekst źródłaStreszczenie:
碩士國立臺北科技大學
商業自動化與管理研究所
91
Artificial neural network (ANN) is machine learning technology. And back-propagation neural network (BPN) is one of the most popular learning paradigms of the network models.Despite the wide application and powerful capacity, A large number of inputs also dramatically increase the computational cost in learning phase. In this paper, we simplify the process of variable extraction and interpret the relative importance between variables. On the basis of back-propagation neural network (BPN), our improvement model which exchanged the input layer for output layer decreases convergent time, so that input variables can be extracted without training first in full model. In addition, we can extract approximate variables with variable extract model. . In conclusion, the accuracy between full model and reduced model which we develop is very close, and sometimes reduced model is even better than full model. Moreover, variable extraction model has converged quickly though the operating time is longer than full model. There are two contributions in our method: saving operation time and extracting variables effectively.
41
Wu, Chen-Ling, i 吳晨翎. "A Novel Classification Algorithm Using Random Back-Propagation Neural Networks". Thesis, 2016. http://ndltd.ncl.edu.tw/handle/5bc47r.
Pełny tekst źródła
42
CHIEN, WOAN-SHIUAN, i 簡婉軒. "Back-propagation and Convolutional Neural Networks for Arrhythmia Electrocardiogram Classification". Thesis, 2017. http://ndltd.ncl.edu.tw/handle/x8bpw6.
Pełny tekst źródłaStreszczenie:
碩士國立中正大學
電機工程研究所
105
Electrocardiogram discrimination plays a very important role in the clinical diagnosis of heart diseases. A lot of methods for ECG beat classification have been proposed. However, many topics still leave room to be improved. In this paper, we build arrhythmia recognition systems based on back-propagation neural network (BPNN) and convolutional neural network (CNN). We will focus on the accuracy of leave one patient out. In this paper, discrete wavelet transform was used to decompose the signal into different subband components, and then higher order statistics were used to describe the ECG signal to enhance the noise-risisting ability. RR-Interval related features were further included to improve the accuracy. Finally, the beat types would be classified by BPNN and CNN. The proposed method is divided into three parts. First, the beat types are classified by BPNN and the validation method is excluding individual differences. Second, the beat types are classified by BPNN and the validation method is leave one patient out. Third, we discuss the relevance of BPNN and CNN and try to combine their advantages to classify the beat types. The results shows that the accuracy by BPNN with 2-Folds validation method is 95.52%. However, while using the leave one patient out validation, the accuracy is 51.4%. Since the arrhythmia beats are far less in number than the normal beats, this phenomenum causes bias in classification and leads to low accuracy. In order to solve these problems, we proposed a two-level classification scheme by adding reference signals of the test sample, adding the features extracted from CNN, and modifying the initial weight of BPNN intending to improve the accuracy of leave one patient out validation. By using the two-level scheme, the best accuracy can reach 91.89% with leave one patient out validation. Compared with the direct classification, the improvement of accuracy is more than 40%.
43
Kuo, Tian-Chuan, i 郭天川. "The Chip Design of Reconfigurable Back Propagation Neural Network Processor". Thesis, 2009. http://ndltd.ncl.edu.tw/handle/58846944404160998384.
Pełny tekst źródłaStreszczenie:
碩士國立暨南國際大學
電機工程學系
97
This paper presents a reconfigurable back-propagation neural network (BPNN) processor hardware architecture as well as doing non-linear function approximation. The purpose of modulation is to make the network more flexible for different applications. The neural network architecture is written in the from of instructions into the chip through the input / output ports (I / O) to reconfigure the neural network architecture. In the implementation of Network Computing, the multi-processor computing architecture is applied in order to reach the parallel computing capacity, and to reduce the time consumed for the learning cycle of back-propagation neural network. The BPNN chip is synthesized by UMC 90nm cell library, and is worked at 50 MHz. The area of the BPNN chip is about 2.2x2.2 mm2 .
44
Chou, Ying-Chieh, i 周穎傑. "The Application of Back-Propagation Network in E-mail Classification". Thesis, 2008. http://ndltd.ncl.edu.tw/handle/34949186966740850122.
Pełny tekst źródłaStreszczenie:
碩士大同大學
資訊工程學系(所)
96
Because of the popularization of Internet and the speed up of the network, people do the first thing after they turn on the computer is read E-mail. Just like the regular post mailbox in the family, our e-mail mailbox was often finding the traces of spam mail. Too much spam has become the biggest worry from user to receive e-mail. The usage of the E-mail will not stop just because of existence of the spam. But the overflowing of spam let user feel vexed endlessly. This is because not only the long time of receive and read mail but also consume the mind to delete and filter the mail. And a large number of junk emails take up the mailbox space, if we don't clearing up immediately; even the normal mail is unable to receive. The main purpose of our research was to develop an e-mail classification system based on Back-Propagation Network. We adopt the technology of automatic text categorization. We first extracted the important features from mail file. Then we use the Chinese segmentation algorithm to process mail subject and content. We using keyword selection and weighting algorithm to find mail keyword and calculate similarity. Finally, we combine Back-Propagation Network and similarity value to achieve the e-mail classification and automatically filter spam mail. The experimental result shows that the system can accomplish the classification function. We also achieve good recall and precision rate in spam mail filtering. We hope to help users to lighten their burdens to receive mails and to reduce the resources of the network; indeed, we reduced the e-mail processing time, but also decrease the amount of spam.
45
Hsu, Chia-Wei, i 許嘉偉. "Back propagation neural networks for optimal design of slab form". Thesis, 2004. http://ndltd.ncl.edu.tw/handle/72775456627616474442.
Pełny tekst źródłaStreszczenie:
碩士國立高雄第一科技大學
營建工程所
95
The traditional method of form design is based on a structural analysis of the form members, which involves many tedious calculations. On the other hand, although experience-based designs are simpler, they do not address all possible scenarios and may be the cause of some accidents due to form failure. The objective of this research is to develop an alternative method for an efficient and safe form design based on back propagation neural networks. The method uses support spacing data as training data to train a neural network, which are obtained from structural analysis for a large number of conditions of loads, member dimensions, and allowable stresses. Taking slab form design as an example, which involves spacing of joists, stringers, and shores, the research found that the network that has two hidden layers with ten nodes each and is trained with 700 training sets can achieve a mean absolute percent error in testing of 1.86%. Then, for local materials and prices and out of 18 combinations of member sizes with the required support spacing for given design loads the optimum design in terms of lowest total cost is determined using the trial-and-error method. A slab form design case is presented to illustrate the proposed approach, in which a saving of 18.7% in total cost can be achieved.
46
Chen, Pen-Chou, i 陳本洲. "Design of Uniaxially Loaded RC ColumnsUsing Back Propagation Neural Network". Thesis, 2007. http://ndltd.ncl.edu.tw/handle/31804969325238018616.
Pełny tekst źródła
47
Mu-Lan, Fan, i 范牧蘭. "A Back-Propagation Networks System For Hospital Patient Churn Premonitory". Thesis, 2004. http://ndltd.ncl.edu.tw/handle/13194404948085332788.
Pełny tekst źródłaStreszczenie:
碩士臺中健康暨管理學院
經營管理研究所
92
The current increasingly competitive pressure has been created in the health medical industry due to the facts such as the performance of health insurance for the entire people, extremely high medical costs, and fast and quick media propagation. It has been taken under consideration by heath medical institutions how to invest in loyal patients with the same health medical concept by using the limited health medical resource to hold existing patients’ loyalty and prevent from potential patients’ churns。 This research is to discuss and analyze the customers’ churn in the issue of customers relationship management for An-Xin clinic, Tai-Pin City, Taizhong County, by means of Artificial Neural Networks, ANNs to construct the analysis mode of customers churns and hope to forecast relevant influential variables for customers’ churns, giving a premonition to health medical institutes prior to customers’ churns. The result of research is as follows: 1.Sexual variables of patients’ are the most significant, among which males’ churn condition is higher than females. 2.Types of illness are secondary, among which the churns of patients with acute diseases are more than ones with chronic diseases. 3.The third is the status of patients, among which the churns of patients at one’s own expense are less than the ones with the states of healthcare insurance. 4.The fourth is marriage state, among which churns of the unmarried are more than the married. 5.The fifth is the age of patients, among which the older are more than the young. Therefore, the management should accurately make policy, adjust the operation of medical treatment and further after learning the churn characteristics. The premonition mode offered by our graduate school can effectively find potential churn patients and possess value of practical applications. Key Words: Foundational Health Treatment, Customer Relationship Management, Customer Churn, Data Mining, Back- Propagation Networks
48
Chang, Po Nien, i 張柏年. "The Automated Wafer Defect Detection based on Back-Propagation Network". Thesis, 2003. http://ndltd.ncl.edu.tw/handle/38406688334990623594.
Pełny tekst źródłaStreszczenie:
碩士國立清華大學
工業工程與工程管理學系
91
For semiconductor industry, the complicated chips manufacturing process, expensive raw materials and demanding production environment make the cost relatively high. Every fabrication company eagerly searches through different manufacturing control and analysis methods to reach the goals of production cost reduction and yield improvement. In a fabrication line, the most direct and efficient way to identify an abnormal process during manufacture is by analyzing the defect maps of semiconductor wafers. Hence it is really crucial to construct a defect inspection system for semiconductor chips, which can provide better judgment and response to the detection of faults. The outcome would be an improved yield rate, the timely identification and elimination of malfunctions and eventual cost reduction. Our current research proposed a defect detection system for semiconductor wafers based on Back-Propagation Network .The major purpose of the current study is to evaluate the characteristics of defects on observed defect maps or bin maps and, thus, obtain the most related defect characteristics through the network training process. Our research will show the effect of different data conversion on the defect detection rate and provide a better output model among the various combinations of parameters. Moreover, our detection system can differentiate between the distinguishable and indistinguishable situations. Based on the verification of real-time manufacturing data provided by online engineers, it is proven that our system can correctly diagnose and analyze the defects such that it fulfills the requirements of semiconductor industry. This system could help engineers find out the problems on the fabrication line in a more timely and efficient manner so that engineers can modify the process to prevent similar problems from recurring, which would consequently stabilize the production process and even improve the yield rate.
49
CHEN, CHAO-CHEN, i 陳昭成. "Back-Propagation Neural Network in NTD/USD exchange rate forecast". Thesis, 2013. http://ndltd.ncl.edu.tw/handle/52949149669114773352.
Pełny tekst źródłaStreszczenie:
碩士育達商業科技大學
休閒事業管理系碩士班
101
Due to high degree of freedom in global trade system and rapid flow of international funds between countries, the increasing fluctuation of exchangerates becomes difficult to predict. Furthermore, the fluctuation of exchange rate has a wide and far-reaching effect, which affects the government agencies, large-scale multinational enterprises, and small companies engaged in import and export business. They also include the leisure industries involved in in-bound and out-bound tourism services. Likewise, the loss and risk to the tourism industries due to the exchange rate fluctuation are difficult to predict. It is very desirable to have an accurate forecast tool to predict and manage the trend of exchange rate fluctuation, and thus to take effective hedging measures and reduce loss and risk to the tourism industries. Therefore, the loss reduction and risk management against the exchange rate fluctuation have become important issues for many import and export businesses. Using the Back-propagation Neural Networks as a research tool and incorporating theSupervised Learning Method to adjust connect weights and optimize the mapping solution from input parameters, this study aims to increase the accuracy of predicting exchange rates. Four major categories of variable model that affect the exchange rate are compared with respect to their predicting accuracy. The criteria used to measure the predicting accuracy are based on calculating the Absolute Relative Error (ARE) and Correlation Coefficient between the actual and predicted values. This study intends to provide the most accurate prediction model that can be used as a forecast reference for tourism business to reduce the risk from exchange rate fluctuation. Other potential uses for forecast reference include assessment of personal finance and investment, corporate financial decisions, importers and exporters to hedge foreign exchange risk, and management strategy of government exchange rate.
50
TUAN, MU HO-YI, i 端木和奕. "Taiwan 4thgeneration stocks with Multiple resilient back-propagation neural models". Thesis, 2015. http://ndltd.ncl.edu.tw/handle/78922808896083689933.
Pełny tekst źródłaStreszczenie:
碩士中華大學
資訊管理學系碩士班
103
Stock market investment and financial management tools .Was Taiwan plople long-term use, and the share price reflect the value of the company in the market. 4th Generation technology in recent years by the impact of stock to produce 4G stocks, the stock vulnerable to introduction of new products fluctuate, How to find tomorrow's ups and downs rules from historical data, The use of artificial intelligence to carry out an objective point of view of data mining elect and future stock price related indicators, and to establish a predictive model to provide decision-making. In this study, Are three 4G stocks, for example, collected in 2012 to 2014, dozens of technology information and indicators, Use Classification And Regression Tree screening the most relevant indicators of closing price, Then use Fuzzy Clustering Method to grouping test data into Back-propagation neural network, of the last time this model and to compare the results of econometric models. Classification And Regression Tree，CART filter out the results D (9), ADX (14) have screened out of the three models, it deserves to be Probe, Back-propagation neural network to predict for UMC Accuracy attain 89.86%,and the output value near to actual value, This study presents the model representative of the trend for change has good predictive ability, three stocks accuracy than ARIMA group higher of 20%, Representing multiple resilient back-propagation neural models, At stock market forecast with a given reference value.