应用科学与工程中的神经网络模型与挑战

5星 · 超过95%的资源需积分: 10 196 浏览量更新于2024-07-21 收藏 6.93MB PDF 举报

"《神经网络在应用科学与工程中的应用》是一本深入探讨神经网络理论与实践的教材，由Taylor & Francis Group, LLC出版。本书以理解生物、生态和自然系统中的复杂性与挑战为起点，引导读者进入神经网络的世界。作者对神经网络的起源、基本原理和它们在处理线性数据分析中的角色进行了详尽阐述。第一章"从数据到模型：理解复杂生物、生态和自然系统中的挑战"首先介绍了书籍的整体结构，强调了理解这些领域中的复杂性和所面临的难题。作者通过实例展示，让读者认识到神经网络在探索和模拟这些系统中的关键作用。第二章深入研究"基础神经网络与模型的线性数据分析"。这一部分首先定义了神经网络的基础概念，解释了它们如何作为一种强大的工具，能够模仿人类大脑的信息处理方式。章节中提到，神经网络的灵感源于生物学，特别是从神经元的工作机制出发，如阈值神经元作为简单的分类器，展示了其基本的决策逻辑。学习策略是本章的核心内容，包括几种常见的学习模型：Hebbian学习，它是一种基于经验和突触强度变化的自组织学习方式；无监督或竞争性学习，强调了网络在没有明确指导的情况下自我组织的能力；以及监督学习，这是一种通过外部反馈进行训练的方法，如经典的感知机模型，用于解决分类和回归问题。作者通过详尽的解释和实际案例，帮助读者建立起对神经网络工作原理的深刻理解，并展示了它们在各种工程和科学研究领域的广泛应用。每一章末尾通常会有参考文献，以便读者进一步深化研究。本书不仅是理论讲解，更是实践指导，适合对神经网络感兴趣的专业人士和学生深入学习。"

9.3.2.1: Single-Step Forecasting with Neural

NARx Model 454

9.3.2.2: Multistep Forecasting with Neural

NARx Model 455

9.3.3: Case Study: River Flow Forecasting 457

9.3.3.1: Linear Model for River Flow Forecasting 460

9.3.3.2: Nonlinear Neural (NARx) Model for River

Flow Forecasting 463

9.3.3.3: Input Sensitivity 467

9.4: Hybrid Linear (ARIMA) and Nonlinear Neural Network Models 468

9.4.1: Case Study: Forecasting the Annual Number of Sunspots 470

9.5: Automatic Generation of Network Structure Using

Simplest Structure Concept 471

9.5.1: Case Study: Forecasting Air Pollution with Automatic

Neural Network Model Generation 473

9.6: Generalized Neuron Network 475

9.6.1: Case Study: Short-Term Load Forecasting with a

Generalized Neuron Network 482

9.7: Dynamically Driven Recurrent Networks 485

9.7.1: Recurrent Networks with Hidden Neuron Feedback 485

9.7.1.1: Encapsulating Long-Term Memory 485

9.7.1.2: Structure and Operation of the Elman Network 488

9.7.1.3: Training Recurrent Networks 490

9.7.1.4: Network Training Example: Hand Calculation 495

9.7.1.5: Recurrent Learning Network Application

Case Study: Rainfall Runoff Modeling 500

9.7.1.6: Two-Step-Ahead Forecasting with Recurrent

Networks 503

9.7.1.7: Real-Time Recurrent Learning Case Study:

Two-Step-Ahead Stream Flow Forecasting 505

9.7.2: Recurrent Networks with Output Feedback 508

9.7.2.1: Encapsulating Long-Term Memory in

Recurrent Networks with Output Feedback 508

9.7.2.2: Application of a Recurrent Net with

Output and Error Feedback and Exogenous

Inputs: (NARIMAx) Case Study: Short-Term

Temperature Forecasting 510

9.7.2.3: Training of Recurrent Nets with

Output Feedback 513

9.7.3: Fully Recurrent Network 515

9.7.3.1: Fully Recurrent Network Practical

Application Case Study: Short-Term Electricity

Load Forecasting 517

9.8: Bias and Variance in Time-Series Forecasting 519

9.8.1: Decomposition of Total Error into Bias and

Variance Components 521

9.8.2: Example Illustrating Bias–Variance Decomposition 522

q 2006 by Taylor & Francis Group, LLC

presented. Many scientists are interested in determining the advantage of

neural networks over classical statistical methods. In this book, statistical

methods are addressed in detail in relation to neural networks; Chapter 2

illustrates that the two approaches are equivalent in linear data analysis and

then begins to build a solid foundation of basic neural network concepts,

instilling a deep understanding to continue forth with conﬁdence.

Chapter 3 through Chapter 5 address nonlinear data analysis with neural

networks using multilayer networks that are the most popular networks for

nonlinear pattern recognition. Multilayer networks are a class of networks

that have layers of neurons with nonlinear processing capabilities. The book

provides extensive coverage of these networks because their potential and

usefulness in systems modeling are increased if their limitations in relation

to robustness and extensive trial-and-error requirements are addressed. The

advantages of neural networks over statistical methods in nonlinear

modeling are illustrated in these chapters. Speciﬁcally, these chapters

address in detail nonlinear processing in networks, network training, and

optimization of network conﬁgurations. Examples and case studies are

presented so that these chapters can be easily understood. The material in

these chapters is intended for both regular lectures and independent study.

Chapter 6 is a discussion of data exploration and preprocessing; it

involves a signiﬁcant number of statistical methods, some of which are

available on commonly known statistical programs. The objective of the

chapter is to extract relevant and independent inputs for effective model

development and it can be used in conjunction with hands-on problem

solving on statistical software.

Chapter 7 discusses uncertainty assessment in neural networks and relies

heavily on statistical methods; neural networks are examined on a rigorous

statistical foundation. Although neural networks are powerful nonlinear

processes, tools to assess their robustness have been limited. In this chapter,

neural networks are put into the context of Bayesian statistics for a rigorous

assessment of their uncertainty in relation to network parameters, errors,

and sensitivities. The material presented in this chapter requires a basic

understanding of the concepts of simple, joint, and conditional probabil-

ities, as well as the neural networks concepts developed in Chapter 3

through Chapter 5. Uncertainty assessment presented in the chapter can be

invaluable for gaining conﬁdence in the neural network models and then

using them in decision making.

In my experience, students are particularly interested in self-organizing

maps—unsupervised networks for discovering unknown clusters and

relationships in multidimensional data. Chapter 6 presents this material in

a step-by-step manner that highlights the important concepts. These can be

used as both lecture material and for independent study. The essential

concepts are presented incrementally and many features of unsupervised

data clustering and its relation to some statistical clustering methods are

xviii

q 2006 by Taylor & Francis Group, LLC

illustrated using examples. Speciﬁcally, the chapter covers topics including

competitive learning and topology preservation, one- and two-dimensional

maps, map training and validation, map quality assessment, cluster

formation on maps, and evolving self-organizing maps using extensive

graphical illustrations.

The last chapter treats linear and nonlinear time-series forecasting with

neural networks. It extensively covers concepts of recurrent and

feedforward networks for short-term and long-term time-series forecasting,

and the majority of the material can be used as both lecture material and for

independent study. A variety of practical example case studies highlight

all new concepts introduced. The similarity of linear neural networks and

the relevant classical statistical methods are illustrated and the advantages of

nonlinear neural networks are demonstrated.

The examples presented in the book have been developed mainly on

Neural Networks for Mathematica

and Machine Learning Framework for

Mathematica—two Mathematica add-on programs—and NeuroShell2e,a

commercial software. There are many commercial and free neural networks

software programs available on the World Wide Web to complement the

material in the book.

It was my intention to present the material in this book in such a way that

the fundamentals gained from it will help the reader apply this knowledge

and understanding to a variety of other networks that are not covered in the

book, as well as any other new developments in this fast-growing ﬁeld. My

experience has shown that the approach used in this book has helped many

diverse researchers learn and apply neural networks in their individual

ﬁelds of research. I hope that the readers will ﬁnd this to be true for

themselves as well.

How to Use the Book

This book covers a number of important issues in model development with

neural networks and is suitable as a research-focused textbook or as a

reference for researchers interested in independent study. The book has

been written for applied scientists and engineers, and as a textbook for

students in these ﬁelds. The material may be presented over two semesters:

Chapter 2 through Chapter 5 may be covered in the ﬁrst semester, and

Chapter 6 through Chapter 9 in the second. Although there is a seamless and

logical progression of material from Chapter 2 through Chapter 7, Chapter 8

relies on the concepts developed in Chapter 2. The ﬁnal chapter has a strong

relationship to Chapter 3 through Chapter 7. I have used NeuroShell2

(Ward Systems, Inc., USA), Neural Connnectione (SPSS, Inc., USA), and

NeuroSolutionse (NeuroDimension, Inc.) in the past and each is suitable for

class tutorials as well as independent research.

xix

q 2006 by Taylor & Francis Group, LLC

剩余581页未读，继续阅读

Shely2017

粉丝: 11
资源: 27

应用科学与工程中的神经网络模型与挑战

Neural Networks for Applied Sciences and Engineering

NeuralNetworks:带有学习算法的简单神经网络项目

no module named 'neural_renderer

No module named 'sklearn.neural_networks'

No module named 'sklearn.neural_networks'咋办

AttributeError: 'bp_neural_network' object has no attribute 'train'

Permission denied: 'C:\\Users\\dell\\Desktop\\QL\\neural_style_transfer\\neural_style_transfer\\png'

AttributeError: module 'neural_renderer' has no attribute 'Renderer'

File "C:\Users\Administrator\AppData\Local\Temp\ipykernel_16012\234815587.py", line 127 def train_and_evaluate_neural_network(num_hidden_layers): ^ IndentationError: unexpected indent

最新资源