Muutke küpsiste eelistusi

E-raamat: Neural Network Analysis, Architectures and Applications

Edited by
  • Formaat: 278 pages
  • Ilmumisaeg: 11-Dec-2024
  • Kirjastus: Institute of Physics Publishing
  • Keel: eng
  • ISBN-13: 9781040287101
Teised raamatud teemal:
  • Formaat - PDF+DRM
  • Hind: 64,99 €*
  • * hind on lõplik, st. muud allahindlused enam ei rakendu
  • Lisa ostukorvi
  • Lisa soovinimekirja
  • See e-raamat on mõeldud ainult isiklikuks kasutamiseks. E-raamatuid ei saa tagastada.
  • Raamatukogudele
Neural Network Analysis, Architectures and ApplicationsSuurem pilt
  • Formaat: 278 pages
  • Ilmumisaeg: 11-Dec-2024
  • Kirjastus: Institute of Physics Publishing
  • Keel: eng
  • ISBN-13: 9781040287101
Teised raamatud teemal:

DRM piirangud

  • Kopeerimine (copy/paste):

    ei ole lubatud

  • Printimine:

    ei ole lubatud

  • Kasutamine:

    Digitaalõiguste kaitse (DRM)
    Kirjastus on väljastanud selle e-raamatu krüpteeritud kujul, mis tähendab, et selle lugemiseks peate installeerima spetsiaalse tarkvara. Samuti peate looma endale  Adobe ID Rohkem infot siin. E-raamatut saab lugeda 1 kasutaja ning alla laadida kuni 6'de seadmesse (kõik autoriseeritud sama Adobe ID-ga).

    Vajalik tarkvara
    Mobiilsetes seadmetes (telefon või tahvelarvuti) lugemiseks peate installeerima selle tasuta rakenduse: PocketBook Reader (iOS / Android)

    PC või Mac seadmes lugemiseks peate installima Adobe Digital Editionsi (Seeon tasuta rakendus spetsiaalselt e-raamatute lugemiseks. Seda ei tohi segamini ajada Adober Reader'iga, mis tõenäoliselt on juba teie arvutisse installeeritud )

    Seda e-raamatut ei saa lugeda Amazon Kindle's. 

Assuming some previous knowledge of neural networks and an understanding of their basic mechanisms, sections describe methods for extracting information about what a neural network has learned, and a method for simplifying network architectures; two novel hardware implementations, together with a discussion of fast training algorithms and the use of several networks in parallel; and three interesting applications: neuro-fuzzy control, data compression, and the use of recurrent networks for target identification. Annotation c. by Book News, Inc., Portland, Or.

This book discusses the main areas of neural networks, with each authoritative chapter covering the latest information from different perspectives. The book first lays the groundwork for understanding and simplifying networks. It then describes novel architectures and algorithms, including pulse-stream techniques, cellular neural networks, and multiversion neural computing. The book concludes by examining various neural network applications, such as neuron-fuzzy control systems and image compression, and presents a case study involving oil spill detection. This book is invaluable for students and practitioners who have a basic understanding of neural computing yet want to broaden and deepen their knowledge of the field.

Neural Network Analysis, Architectures and Applications discusses the main areas of neural networks, with each authoritative chapter covering the latest information from different perspectives. Divided into three parts, the book first lays the groundwork for understanding and simplifying networks. It then describes novel architectures and algorithms, including pulse-stream techniques, cellular neural networks, and multiversion neural computing. The book concludes by examining various neural network applications, such as neuron-fuzzy control systems and image compression. This final part of the book also provides a case study involving oil spill detection. This book is invaluable for students and practitioners who have a basic understanding of neural computing yet want to broaden and deepen their knowledge of the field.
Preface xiii
PART 1 Understanding and Simplifying Networks 1(100)
1 Analyzing the Internal Representations of Trained Neural Networks
3(24)
1.1 Introduction
3(2)
1.2 Analysis techniques
5(19)
1.2.1 Hierarchical cluster analysis
6(4)
1.2.2 Principal component analysis
10(1)
1.2.3 Multi-dimensional scaling
11(2)
1.2.4 Discriminant analysis
13(5)
1.2.5 Output weight projections
18(5)
1.2.6 Contribution analysis
23(1)
1.3 Conclusions
24(3)
2 Information Maximization to Simplify Internal Representation
27(34)
2.1 Introduction
27(1)
2.2 Information maximization method
28(6)
2.2.1 Outline of information maximization
28(1)
2.2.2 The concept of information
28(2)
2.2.3 An information theoretic formulation
30(1)
2.2.4 Information maximization
31(2)
2.2.5 Cross entropy minimization
33(1)
2.3 Modified information maximization methods
34(4)
2.3.1 Outline of modified methods
34(1)
2.3.2 Information maximizer
35(2)
2.3.3 Information maximizer with weigth elimination
37(1)
2.4 Application to the XOR problem
38(3)
2.5 Application to the symmetry problem
41(3)
2.6 Application to well-formedness estimation
44(13)
2.6.1 Well-formedness by sonority
44(1)
2.6.2 Information and generalization
45(1)
2.6.3 Interpretation of internal representation
46(5)
2.6.4 Improved generalization with weight elimination
51(6)
2.7 Competitive learning and information maximization
57(2)
2.8 Conclusion
59(2)
3 Rule Extraction From Trained Artificial Neural Networks
61(40)
3.1 Introduction
61(1)
3.2 The importance of rule extraction
62(4)
3.2.1 Provision of a `user explanation' capability
63(1)
3.2.2 Extension of ANN systems to `safety-critical' problem domains
64(1)
3.2.3 Software verification and debugging of ANN components in software systems
64(1)
3.2.4 Improving the generalization of ANN solutions
65(1)
3.2.5 Data exploration and the induction of scientific theories
65(1)
3.2.6 Knowledge acquisition for symbolic AI systems
65(1)
3.3 Problem overview
66(1)
3.4 A classification scheme for rule-extraction algorithms
67(3)
3.4.1 The expressive power of the extracted rules
67(1)
3.4.2 The translucency of the underlying ANN units
67(2)
3.4.3 The portability of the rule-extraction algorithm
69(1)
3.4.4 Quality of the extracted rules
69(1)
3.4.5 Complexity of the rule-extraction algorithm
70(1)
3.5 Rule-extraction techniques
70(12)
3.5.1 Decompositional approaches to rule extraction
70(2)
3.5.2 Decompositional algorithms that directly decompile weights to rules
72(5)
3.5.3 Pedagogical approaches to rule extraction
77(1)
3.5.4 The VIA algorithm
78(3)
3.5.5 Eclectic rule-extraction techniques
81(1)
3.6 Extraction of fuzzy rules
82(2)
3.7 Techniques for performing rule refinement
84(4)
3.8 Rule refinement and recurrent networks
88(4)
3.9 Current issues in rule extraction and refinement
92(6)
3.9.1 Limitations imposed by inherent algorithmic complexity
92(1)
3.9.2 Limitations on achieving simultaneously high accuracy, high fidelity, and high comprehensibility
93(1)
3.9.3 Rule extraction and the quality of ANN solutions
94(1)
3.9.4 Functional dependences, causal factors and rule extraction
95(3)
3.9.5 Extension to connectionist knowledge representation techniques
98(1)
3.10 Conclusion
98(3)
PART 2 Novel Architectures and Algorithms 101(82)
4 Pulse-Stream Techniques and Circuits for Implementing Neural Networks
103(7)
4.1 Introduction
103(1)
4.2 Pulse-stream encoding
103(1)
4.3 Basic neural computation on a chip
104(2)
4.4 Computation using an analogue, two-quadrant multiplier
106(2)
4.5 Designing a pulsed multiplier-a case study
108(3)
4.6 Implementing weights
111(1)
4.7 Results from the multiplier circuit
112(2)
4.8 Converting analogue outputs and inputs into pulses
114(1)
4.9 Centred pulses
115(1)
4.10 Quantization
115(2)
4.11 Process dependence of signals as VLSI technologies are scaled down
117(1)
4.12 ANN, and other, applications
118(1)
4.13 Further reading on applications of the technique
118(2)
5 Cellular Neural Networks
120(16)
5.1 Introduction
120(1)
5.2 The CNN paradigm
121(6)
5.3 CNN dynamical systems
127(7)
5.4 Conclusions: the CNN as supercomputer or neural network?
134(2)
6 Efficient Training of Feed-Forward Neural Networks
136(38)
6.1 Introduction
136(1)
6.2 Notation and basic definitions
137(1)
6.3 Optimization strategy
138(1)
6.4 Gradient descent
139(13)
6.4.1 Back-propagation
140(1)
6.4.2 Convergence rate
141(2)
6.4.3 Gradient descent with momentum
143(1)
6.4.4 Adaptive learning rate and momentum
144(3)
6.4.5 Learning rate schedules for on-line gradient descent
147(1)
6.4.6 The quickprop method
148(1)
6.4.7 Estimation of optimal learning rate and reduction of large curvature components
149(3)
6.5 Conjugate gradient
152(13)
6.5.1 Non-interfering directions of search
152(3)
6.5.2 Convergence rate
155(2)
6.5.3 Scaled conjugate gradient
157(6)
6.5.4 Stochastic conjugate gradient
163(2)
6.6 Newton related methods
165(1)
6.7 On-line versus off-line discussion
166(6)
6.8 Conclusion
172(2)
7 Exploiting Local Optima in Multiversion Neural Computing
174(9)
7.1 Local optima as a neural computing problem
174(1)
7.2 The group properties of local optima
175(2)
7.3 Engineering productive sets of local optima
177(1)
7.4 Optimizing performance over local optima sets
177(2)
7.5 Applications of the multiversion approach
179(1)
7.6 Discussion and conclusions
180(3)
PART 3 Applications 183(68)
8 Neural and Neuro-Fuzzy Control Systems
185(19)
8.1 Introduction
185(1)
8.2 Traditional control
185(2)
8.3 State space
187(1)
8.4 Neural control
188(5)
8.4.1 Approximating the inverse transfer function
188(1)
8.4.2 Emulating a controller
189(1)
8.4.3 Reinforcement learning
190(3)
8.5 Fuzzy control
193(4)
8.5.1 Background
193(1)
8.5.2 Fuzzy logic
193(1)
8.5.3 Fuzzy sets and membership functions
193(1)
8.5.4 Fuzzy logic control (FLC)
194(3)
8.6 Neuro-fuzzy control
197(6)
8.6.1 Adaptive fuzzy associative memory (AFAM)
197(3)
8.6.2 Takagi-Sugeno-Kang (TSK)
200(1)
8.6.3 B-spline functions
201(2)
8.7 Summary
203(1)
9 Image Compression using Neural Networks
204(20)
9.1 Introduction
204(1)
9.2 Metrics
205(1)
9.3 Basic image compression
206(8)
9.3.1 Basic compression implementations
208(2)
9.3.2 CODECs with multiple compression networks
210(1)
9.3.3 Adaptive methods
210(2)
9.3.4 Different neural network models
212(2)
9.4 The random neural network model
214(4)
9.4.1 Image compression with the RNN
216(2)
9.5 Vector quantization
218(4)
9.5.1 Adaptive vector quantization
219(1)
9.5.2 Adaptive Kohonen networks
220(1)
9.5.3 VLSI Kohonen networks
220(1)
9.5.4 Predictive vector quantization
221(1)
9.6 Conclusions
222(2)
10 Oil Spill Detection: a Case Study using Recurrent Artificial Neural Networks
224(27)
10.1 Introduction
224(1)
10.2 Problem description
225(7)
10.2.1 Overview
225(1)
10.2.2 The SLAR simulation model
225(2)
10.2.3 Data analysis
227(3)
10.2.4 Design of network architecture
230(2)
10.3 Experiments
232(9)
10.3.1 The architectures
232(1)
10.3.2 Reliability and sensitivity
233(4)
10.3.3 Robustness to varying sea states
237(2)
10.3.4 Further experiments
239(2)
10.4 Analysis and discussion
241(7)
10.4.1 Functional analysis
242(2)
10.4.2 Analysis of internal states
244(4)
10.5 Summary and conclusion
248(3)
Bibliography 251(12)
Index 263
Lisainfo e-raamatute kohta Lisainfo e-raamatute kohta
Püsilink: https://www.kriso.ee/db/97810402871012e.html
Märksõnad: