DOCSLIB.ORG

An Introduction to Neural Networks

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
An Introduction to Neural Networks An introduction to Neural Networks .. Ben Krose Patrick van der Smagt Eighth edition November c The Universityof Amsterdam Permission is granted to distribute single copies of this book for noncommercial use as long as it is distributed as a whole in its original form and the names of the authors and the University of Amsterdam are mentioned Permission is also granted to use this book for noncommercial courses provided the authors are notied of this b eforehand The authors can b e reached at Ben Krose Patrickvan der Smagt Faculty of Mathematics Computer Science Institute of Rob otics and System Dynamics University of Amsterdam German Aerospace Research Establishment Kruislaan NL SJ Amsterdam POBox D Wessling THE NETHERLANDS GERMANY Phone Phone Fax Fax email krosefwiuvanl email smagtdlrde URL httpwwwfwiuvanlresearchneuro URL httpwwwopdlrdeFFDRRS Contents Preface I FUNDAMENTALS Intro duction Fundamentals A framework for distributed representation Pro cessing units Connections b etween units Activation and output rules Network top ologies Training of articial neural networks Paradigms of learning Mo difying patterns of connectivity Notation and terminology Notation Terminology II THEORY Perceptron and Adaline Networks with threshold activation functions Perceptron learning rule and convergence theorem Example of the Perceptron learning rule Convergence theorem The original Perceptron The adaptive linear element Adaline Networks with linear activation functions the delta rule ExclusiveOR problem Multilayer p erceptrons can do everything Conclusions BackPropagation Multilayer feedforward networks The generalised delta rule Understanding backpropagation Working with backpropagation An example Other activation functions CONTENTS Deciencies of backpropagation Advanced algorithms How go o d are multilayer feedforward networks The eect of the numb er of learning samples The eect of the numb er of hidden units Applications Recurrent Networks The generalised deltarule in recurrentnetworks The Jordan network The Elman network Backpropagation in fully recurrentnetworks The Hopeld network Description Hopeld network as asso ciative memory Neurons with graded resp onse Boltzmann machines SelfOrganising Networks Comp etitive learning Clustering Vector quantisation Kohonen network Principal comp onent networks Intro duction Normalised Hebbian rule Principal comp onent extractor More eigenvectors Adaptive resonance theory Background Adaptive resonance theory ART The simplied neural network mo del ART The original mo del Reinforcement learning The critic The controller network Bartos approach the ASEACE combination Asso ciativesearch Adaptive critic The cartp ole system Reinforcementlearningversus optimal control III APPLICATIONS Rob ot Control Endeector p ositioning Camerarob ot co ordination is function approximation Rob ot arm dynamics Mobile rob ots Mo del based navigation Sensor based control CONTENTS Vision Intro duction Feedforward typ es of networks Selforganising networks for image compression Backpropagation Linear networks Principal comp onents as features The cognitron and neo cognitron Description of the cells Structure of the cognitron Simulation results Relaxation typ es of networks Depth from stereo Image restoration and image segmentation Silicon retina IV IMPLEMENTATIONS General Purp ose Hardware The Connection Machine Architecture Applicability to neural networks Systolic arrays Dedicated NeuroHardware General issues Connectivity constraints Analogue vs digital Optics Learning vs nonlearning Implementation examples Carver Meads silicon retina LEPs LNeuro chip References Index CONTENTS List of Figures The basic comp onents of an articial neural network Various activation functions for a unit Single layer network with one output and two inputs Geometric representation of the discriminant function and the weights Discriminant function b efore and after weight up date The Perceptron The Adaline .
Load more

Recommended publications
  • A Adaboost, 176 ADALINE. See Adaptive Linear Element/ Adaptive
    Index A B AdaBoost, 176 Bayes’ classifier, 171–173 ADALINE. See Adaptive Linear Element/ Bayesian decision theory Adaptive Linear Neuron multiple features (ADALINE) complex decision boundary, Adaptive Linear Element/Adaptive Linear 85, 86 Neuron (ADALINE), 112 trade off performance, 86 Agglomerative hierarchical clustering. See two-dimensional feature space, 85 Hierarchical clustering single dimensional (1D) feature Angular anisotropy, 183 class-conditional probabilities (see ANN. See Artificial neural network (ANN) Class-conditional probabilities) Artificial neural network (ANN) classification error, 81–83 activation function, 112, 113 CondProb.xls, 85 ADALINE, 112 likelihood ratio, 77, 83, 84 advantages and disadvantages, 116, 117 optimal decision rule, 81 backpropagation, 109, 115 posterior probability, 76–77 bias, 99–100, 105–106, 109 recognition approaches, 75 feed-forward, 112, 113 symmetrical/zero-one loss global minimum, 110 function, 83 learning rate, 109–111 Bayes’ rule linear discriminant function, 107 conditional probability and, 46–53 logical AND function, 106, 107 Let’s Make a Deal, 48 logical OR function, 107 naı¨ve Bayes classifier, 53–54 logical XOR function, 107–109, 114 posterior probability, 47 multi-layer network, 107, 112, 116 Bias-variance tradeoff, 99–101 neurons, 104–105 Bioinformatics, 2 overfitting, 112–113 Biometrics, 2 perceptron (see Perceptron, ANN) Biometric systems recurrent network, 112 cost, accuracy, and security, 184 signals, 104 face recognition, 187 simulated annealing, 115 fingerprint recognition, 184–186 steepest/gradient descent, 110, 115 infrared scan, hand, 184 structure, neurons, 105 physiological/behavioral training, 109, 115–116 characteristics, 183 universal approximation theorem, 112 Bootstrap, 163–164 weight vector, 109, 112 Breast Screening, 51–52 Astronomy, 2 Brodatz textures, 181 G.
    [Show full text]
  • Chapter 02: Fundamentals of Neural Networks
    FUNDAMENTALS OF NEURAL 2 NETWORKS Ali Zilouchian 2.1 INTRODUCTION For many decades, it has been a goal of science and engineering to develop intelligent machines with a large number of simple elements. References to this subject can be found in the scientific literature of the 19th century. During the 1940s, researchers desiring to duplicate the function of the human brain, have developed simple hardware (and later software) models of biological neurons and their interaction systems. McCulloch and Pitts [1] published the first systematic study of the artificial neural network. Four years later, the same authors explored network paradigms for pattern recognition using a single layer perceptron [2]. In the 1950s and 1960s, a group of researchers combined these biological and psychological insights to produce the first artificial neural network (ANN) [3,4]. Initially implemented as electronic circuits, they were later converted into a more flexible medium of computer simulation. However, researchers such as Minsky and Papert [5] later challenged these works. They strongly believed that intelligence systems are essentially symbol processing of the kind readily modeled on the Von Neumann computer. For a variety of reasons, the symbolic–processing approach became the dominant method. Moreover, the perceptron as proposed by Rosenblatt turned out to be more limited than first expected. [4]. Although further investigations in ANN continued during the 1970s by several pioneer researchers such as Grossberg, Kohonen, Widrow, and others, their works received relatively less attention. The primary factors for the recent resurgence of interest in the area of neural networks are the extension of Rosenblatt, Widrow and Hoff’s works dealing with learning in a complex, multi-layer network, Hopfield mathematical foundation for understanding the dynamics of an important class of networks, as well as much faster computers than those of 50s and 60s.
    [Show full text]
  • Investigation of Stochastic Gradient Descent in Neural Networks
    Investigation of Stochastic Gradient Descent in Neural Networks Hugh Chen Department of Statistics University of Washington Seattle, Washington {hughchen}@uw.edu Abstract Neural networks are machine learning techniques that are capable of arbitrarily reducing bias by introducing parameters and non-linearities - which often leads to non-convex optimization. Due to the generalizability and widespread appeal of neural networks, many smart techniques have been developed to optimize neural networks. In this paper, I investigate and specify different variants of stochastic gradient descent for neural networks. Then, I explore these methods experimentally. First, I experiment with methods for visualizing these methods for gradient descent. Second, I analyze these methods applied to neural networks in simple settings. Finally, I look at the utility of ensemble methods which can be applied in addition to these variants of SGD. Keywords: neural networks, stochastic optimization, ensembles. 1 Background Stochastic gradient descent (SGD) is a stochastic approximation for minimizing an objective function (generally a loss function) [1]. Defining a function L over a set of weight parameters w, our general form of objective function is: n X L(w; y1; ··· ; yn; x1; ··· ; xn) = Li(w; yi; xi) i=1 We denote a prediction using weights w on input xi as h(w; xi). For regression, mean squared error is a commonly used loss function where 2 Li(w; yi; xi) = (yi − h(w; xi)) For binary classification problems, it is common to utilize binary cross-entropy loss functions where Li(w; yi; xi) = yi ln(h(w; xi)) + (1 − yi) ln(1 − h(w; xi)) Other loss functions include exponential cost, Hellinger distance, Kullback-Leibler divergence, generalized Kullback-Leibler divergence, and Itakura-Saito distance [2].
    [Show full text]
  • ADALINE Approach for Induction Motor Mechanical Parameters Identification Hamid Sediki, Ali Bechouche, Djaffar Ould Abdeslam, Salah Haddad
    ADALINE approach for induction motor mechanical parameters identification Hamid Sediki, Ali Bechouche, Djaffar Ould Abdeslam, Salah Haddad To cite this version: Hamid Sediki, Ali Bechouche, Djaffar Ould Abdeslam, Salah Haddad. ADALINE approach for induc- tion motor mechanical parameters identification. Mathematics and Computers in Simulation, Elsevier, 2013, 90 (-), pp.86-97. 10.1016/j.matcom.2012.05.003. hal-00985414 HAL Id: hal-00985414 https://hal.archives-ouvertes.fr/hal-00985414 Submitted on 29 Apr 2014 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. 1 ADALINE approach for induction motor mechanical parameters identification Hamid Sediki a, Ali Bechouche a, Djaffar Ould Abdeslam b,*, Salah Haddad a a Department of Electrical Engineering, Mouloud Mammeri University, BP 17 RP Tizi-Ouzou, Algeria b MIPS Laboratory, University of Haute-Alsace, 4 rue des Frères Lumières, 68093 Mulhouse, France * Corresponding author. Tel.: 33 (0)3 89 33 60 20 E-mail address: [email protected] (D. Ould Abdeslam). Abstract Two new methods to identify the mechanical parameters in induction motor based field oriented drives are presented in this paper. The identified parameters are: the moment of inertia and the viscous damping coefficient.
    [Show full text]
  • Neural Networks Learning the Network: Part 1
    Neural Networks Learning the network: Part 1 11-785, Spring 2018 Lecture 3 1 Designing a net.. • Input: An N-D real vector • Output: A class (binary classification) • “Input units”? • Output units? • Architecture? • Output activation? 2 Designing a net.. • Input: An N-D real vector • Output: Multi-class classification • “Input units”? • Output units? • Architecture? • Output activation? 3 Designing a net.. • Input: An N-D real vector • Output: Real-valued output • “Input units”? • Output units? • Architecture? • Output activation? 4 Designing a net.. • Conversion of real number to binary representation – Input: A real number – Output: The binary sequence for the number • “Input units”? • Output units? • Architecture? • Output activation? 5 activatation 1 -1 1/-1 w 6 activatation 1 -1 1/-1 1/-1 w/2 w w X 7 activatation 1 -1 1/-1 1/-1 1/-1 w/2 w/4 w w/2 w w X 8 Designing a net.. • Binary addition: – Input: Two binary inputs – Output: The binary (bit-sequence) sum • “Input units”? • Output units? • Architecture? • Output activation? 9 Designing a net.. • Clustering: – Input: Real-valued vector – Output: Cluster ID • “Input units”? • Output units? • Architecture? • Output activation? 10 Topics for the day • The problem of learning • The perceptron rule for perceptrons – And its inapplicability to multi-layer perceptrons • Greedy solutions for classification networks: ADALINE and MADALINE • Learning through Empirical Risk Minimization • Intro to function optimization and gradient descent 11 Recap • Neural networks are universal function approximators
    [Show full text]
  • Simulations of Artificial Neural Network with Memristive Devices
    SIMULATIONS OF ARTIFICIAL NEURAL NETWORK WITH MEMRISTIVE DEVICES by Thanh Thi Thanh Tran A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Electrical Engineering Boise State University December 2012 c 2012 Thanh Thi Thanh Tran ALL RIGHTS RESERVED BOISE STATE UNIVERSITY GRADUATE COLLEGE DEFENSE COMMITTEE AND FINAL READING APPROVALS of the thesis submitted by Thanh Thi Thanh Tran Thesis Title: Simulations of Artificial Neural Network with Memristive Devices Date of Final Oral Examination: December 2012 The following individuals read and discussed the thesis submitted by student Thanh Thi Thanh Tran, and they evaluated her presentation and response to questions during the final oral examination. They found that the student passed the final oral examination. Elisa H. Barney Smith, Ph.D. Chair, Supervisory Committee Kristy A. Campbell, Ph.D. Member, Supervisory Committee Vishal Saxena, Ph.D. Member, Supervisory Committee The final reading approval of the thesis was granted by Elisa H. Barney Smith, Ph.D., Chair, Supervisory Committee. The thesis was approved for the Graduate College by John R. Pelton, Ph.D., Dean of the Graduate College. Dedicated to my parents and Owen. iv ACKNOWLEDGMENTS I would like to thank Dr. Barney Smith who always encourages me to keep pushing my limit. This thesis would not have been possible without her devoting support and mentoring. I wish to thank Dr. Saxena for guiding me through some complex circuit designs. I also wish to thank Dr. Campbell for introducing me to the memristive devices and helping me to understand important concepts to be able to program the device.
    [Show full text]
  • Programming Neural Networks with Encog3 in C
    Programming Neural Networks with Encog3 in C# Programming Neural Networks with Encog3 in C# Jeff Heaton Heaton Research, Inc. St. Louis, MO, USA v Publisher: Heaton Research, Inc Programming Neural Networks with Encog 3 in C# First printing October, 2011 Author: Jeff Heaton Editor: WordsRU.com Cover Art: Carrie Spear ISBN’s for all Editions: 978-1-60439-026-1, PAPER 978-1-60439-027-8, PDF 978-1-60439-029-2, NOOK 978-1-60439-030-8, KINDLE Copyright ©2011 by Heaton Research Inc., 1734 Clarkson Rd. #107, Chester- field, MO 63017-4976. World rights reserved. The author(s) created reusable code in this publication expressly for reuse by readers. Heaton Research, Inc. grants readers permission to reuse the code found in this publication or down- loaded from our website so long as (author(s)) are attributed in any application containing the reusable code and the source code itself is never redistributed, posted online by electronic transmission, sold or commercially exploited as a stand-alone product. Aside from this specific exception concerning reusable code, no part of this publication may be stored in a retrieval system, trans- mitted, or reproduced in any way, including, but not limited to photo copy, photograph, magnetic, or other record, without prior agreement and written permission of the publisher. Heaton Research, Encog, the Encog Logo and the Heaton Research logo are all trademarks of Heaton Research, Inc., in the United States and/or other countries. TRADEMARKS: Heaton Research has attempted throughout this book to distinguish proprietary trademarks from descriptive terms by following the capitalization style used by the manufacturer.
    [Show full text]
  • Normas Para Redação Do Trabalho Completo
    Rio de Janeiro, v.6, n.2, p. 299-317, maio a agosto de 2014 DECISÃO COM REDES NEURAIS ARTIFICIAIS EM MODELOS DE SIMULAÇÃO A EVENTOS DISCRETOS Marília Gonçalves Dutra da Silvaab, Joao Jose de Assis Rangela*, David Vasconcelos Corrêa da Silvaab, Túlio Almeida Peixotoa, Ítalo de Oliveira Matiasa aUniversidade Candido Mendes – UCAM, Campos dos Goytacazes – RJ, Brasil bInstituto Federal Fluminense – IFF, Campos dos Goytacazes – RJ, Brasil Resumo O objetivo deste trabalho é avaliar frameworks para aplicação de redes neurais artificiais acoplados a modelos de simulação a eventos discretos com o software Ururau. A análise está centrada em encontrar mecanismos que possibilitem a realização de decisões através de outros algoritmos integrados ao código do modelo de simulação. Na literatura atual, há pouca citação de como aplicar algum tipo de inteligência computacional a partes do código de um modelo, que necessite refletir comportamentos de ações com aspectos de maior complexidade, como ações de pessoas em uma parte do processo. Por outro lado, muitos softwares de simulação permitem a customização dos modelos através de algoritmos que podem ser integrados aos modelos. Os resultados encontrados demonstraram que a utilização do framework ENCOG se mostrou adequado para a criação das redes neurais artificiais, sem apresentar erros de funcionamento e incompatibilidades com o código do software Ururau. Além disso, o trabalho permite que as pessoas interessadas em compreender esta estrutura possam visualizá-la diretamente no código do modelo proposto. Palavras-chave: Ururau, Redes Neurais Artificiais, ENCOG, Decisão. Abstract The objective of this study is to evaluate frameworks for the application of artificial neural networks linked to models of discrete event simulation with the Ururau software.
    [Show full text]
  • A Novel Approach for an MPPT Controller Based on the ADALINE Network Trained with the RTRL Algorithm
    energies Article A Novel Approach for an MPPT Controller Based on the ADALINE Network Trained with the RTRL Algorithm Julie Viloria-Porto *, Carlos Robles-Algarín * and Diego Restrepo-Leal * Facultad de Ingeniería, Universidad del Magdalena, Santa Marta 470003, Colombia * Correspondence: [email protected] (J.V.-P.); [email protected] (C.R.-A.); [email protected] (D.R.-L.); Tel.: +57-5-421-7940 (C.R.-A.) Received: 9 November 2018; Accepted: 27 November 2018; Published: 5 December 2018 Abstract: The Real-Time Recurrent Learning Gradient (RTRL) algorithm is characterized by being an online learning method for training dynamic recurrent neural networks, which makes it ideal for working with non-linear control systems. For this reason, this paper presents the design of a novel Maximum Power Point Tracking (MPPT) controller with an artificial neural network type Adaptive Linear Neuron (ADALINE), with Finite Impulse Response (FIR) architecture, trained with the RTRL algorithm. With this same network architecture, the Least Mean Square (LMS) algorithm was developed to evaluate the results obtained with the RTRL controller and then make comparisons with the Perturb and Observe (P&O) algorithm. This control method receives as input signals the current and voltage of a photovoltaic module under sudden changes in operating conditions. Additionally, the efficiency of the controllers was appraised with a fuzzy controller and a Nonlinear Autoregressive Network with Exogenous Inputs (NARX) controller, which were developed in previous investigations. It was concluded that the RTRL controller with adaptive training has better results, a faster response, and fewer bifurcations due to sudden changes in the input signals, being the ideal control method for systems that require a real-time response.
    [Show full text]
  • Neural Networks : Soft Computing Course Lecture 7 – 14, Notes, Slides , RC Chakraborty, E-Mail [email protected] , Aug
    Fundamentals of Neural Networks : Soft Computing Course Lecture 7 – 14, notes, slides www.myreaders.info/ , RC Chakraborty, e-mail [email protected] , Aug. 10, 2010 http://www.myreaders.info/html/soft_computing.html www.myreaders.info RC Chakraborty, www.myreaders.info Fundamentals of Neural Networks Soft Computing Neural network, topics : Introduction, biological neuron model, artificial neuron model, neuron equation. Artificial neuron : basic elements, activation and threshold function, piecewise linear and sigmoidal function. Neural network architectures : single layer feed- forward network, multi layer feed-forward network, recurrent networks. Learning methods in neural networks : unsupervised Learning - Hebbian learning, competitive learning; Supervised learning - stochastic learning, gradient descent learning; Reinforced learning. Taxonomy of neural network systems : popular neural network systems, classification of neural network systems as per learning methods and architecture. Single-layer NN system : single layer perceptron, learning algorithm for training perceptron, linearly separable task, XOR problem, ADAptive LINear Element (ADALINE) - architecture, and training. Applications of neural networks: clustering, classification, pattern recognition, function approximation, prediction systems. Fundamentals of Neural Networks Soft Computing Topics (Lectures 07, 08, 09, 10, 11, 12, 13, 14 8 hours) RC Chakraborty, www.myreaders.info Slides 1. Introduction 03-12 Why neural network ?, Research History, Biological Neuron model, Artificial
    [Show full text]
  • Investigating Signal Power Loss Prediction in a Metropolitan Island Using ADALINE and Multi-Layer Perceptron Back Propagation Networks
    International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13409-13420 © Research India Publications. http://www.ripublication.com Investigating Signal Power Loss Prediction in A Metropolitan Island Using ADALINE and Multi-Layer Perceptron Back Propagation Networks Virginia Chika Ebhota1*, Joseph Isabona2, Viranjay M. Srivastava3 1, 2. 3Department of Electronic Engineering, Howard College, University of KwaZulu-Natal, Durban, South Africa. Correspondence author Abstract or sparsely connected, binary or floating point in operation, asynchronous or synchronous, static or dynamic, discrete or The prediction ability of every artificial neural network model continuous and adopt supervise or unsupervise learning. depends on number of factors such as the architectural However, the effort to make choice among the various ANN complexity of the network, the network parameters, the architectures and establish which is superior likely fails in underlying physical problem, etc. The performance of two most cases, as the choice should be application oriented [1]. It ANN architectures: ADALINE and MLP network in is preferred to test various types of ANNs for every novel predicting the signal power loss during electromagnetic signal application than making use of preselected one that performed propagation using measured data from LTE network was well for a particular problem. Prior to the training of ANNs, it explored in this work. Learning rate parameter was used to the is important that the network architecture is well specified as check the predictive effectiveness of the two network models the learning ability of the network and its performance is while considering the prediction error between the actual dependent on how suitable the architecture is for the case measurement and the prediction values using four under study.
    [Show full text]
  • Technologies
    technologies Perspective Memristors for the Curious Outsiders Francesco Caravelli 1,* and Juan Pablo Carbajal 2,3 1 Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, NM 87545, USA 2 Institute for Energy Technology, HSR (Hochschule für Technik Rapperswil) or University of Applied Sciences Rapperswil, 8640 Rapperswil, Switzerland; [email protected] 3 Swiss Federal Institute of Aquatic Science and Technology, Eawag, Überlandstrasse 133, 8600 Dübendorf, Switzerland * Correspondence: [email protected] Received: 28 September 2018; Accepted: 3 December 2018; Published: 9 December 2018 Abstract: We present both an overview and a perspective of recent experimental advances and proposed new approaches to performing computation using memristors. A memristor is a 2-terminal passive component with a dynamic resistance depending on an internal parameter. We provide an brief historical introduction, as well as an overview over the physical mechanism that lead to memristive behavior. This review is meant to guide nonpractitioners in the field of memristive circuits and their connection to machine learning and neural computation. Keywords: memristors; neuromorphic computing; analog computation; machine learning 1. Introduction The present review aims at providing a structured view over the many areas in which memristor technology is becoming popular. As in many other hyped topics, there is a risk that most of the activity we see today will dissipate into smoke in the coming years. Hence, we have carefully selected a set of topics for which we have experience and we believe will remain relevant when memristors move out of the spotlight. After a general overview on memristors, we provide a historical overview of the topic.
    [Show full text]