Part I |
Basic
Concepts |
|
1 |
Defining the Area |
|
|
1.1 |
Learning from Information |
|
|
1.2 |
General Objectives and Concepts |
|
|
1.3 |
What Neural Networks are Good for |
|
|
1.4 |
Notation, Conventions and Abbreviations |
|
|
1.5 |
Beyond a Printed Edition |
|
|
2 |
Neuron |
|
|
2.1 |
Synapses and Input Signals |
|
|
2.2 |
Weights |
|
|
2.3 |
Linear Learning Machine |
|
|
2.4 |
Transfer Functions in Neurons |
|
|
2.5 |
Bias |
|
|
2.6 |
Graphical Representation of Artificial
Neurons |
|
|
2.7 |
Essentials |
|
|
2.8 |
References and Suggested Readings |
|
|
3 |
Linking Neurons into Networks |
|
|
3.1 |
General |
|
|
3.2 |
One Layer |
|
|
3.3 |
Input |
|
|
3.4 |
Architectures |
|
|
3.5 |
Hidden Layer; Output Layer |
|
|
3.6 |
Graphical Representation of Neural
Networks |
|
|
3.7 |
Essentials |
|
|
3.8 |
References and Suggested Readings |
|
Part II |
One-Layer
Networks |
|
4 |
Hopfield Network |
|
|
4.1 |
General |
|
|
4.2 |
Architecture |
|
|
4.3 |
Transfer Function |
|
|
4.4 |
Weight Matrix |
|
|
4.5 |
Iteration |
|
|
4.6 |
Capacity of the Hopfield Network |
|
|
4.7 |
Essentials |
|
|
4.8 |
References and Suggested Readings |
|
|
5 |
Adaptive Bidirectional Associative
Memory (ABAM) |
|
|
5.1 |
Unsupervised and Supervised Learning |
|
|
5.2 |
General |
|
|
5.3 |
ABAM Network |
|
|
5.4 |
Learning Procedure |
|
|
5.5 |
An Example |
|
|
5.6 |
Significance of the Example |
|
|
5.7 |
Essentials |
|
|
5.8 |
References and Suggested Readings |
|
|
6 |
Kohonen Network |
|
|
6.1 |
General |
|
|
6.2 |
Architecture |
|
|
6.3 |
Competitive Learning |
|
|
6.4 |
Mapping from Three to Two Dimensions |
|
|
6.5 |
Another Example |
|
|
6.6 |
Remarks |
|
|
6.7 |
Essentials |
|
|
6.8 |
References and Suggested Readings |
|
Part III |
Multilayer
Networks |
|
7 |
Counter-Propagation |
|
|
7.1 |
Transition from One to Two Layers |
|
|
7.2 |
Lookup Table |
|
|
7.3 |
Architecture |
|
|
7.4 |
Supervised Competitive Learning |
|
|
7.5 |
Learning to Play Tennis |
|
|
7.6 |
Correlations Among the Variables |
|
|
7.7 |
Essentials |
|
|
7.8 |
References and Suggested Readings |
|
|
8 |
Back-Propagation of Errors |
|
|
8.1 |
General |
|
|
8.2 |
Architecture |
|
|
8.3 |
Learning by Back-Propagation |
|
|
8.4 |
The Generalized Delta-Rule |
|
|
8.5 |
Learning Algorithm |
|
|
8.6 |
Example: Tennis Match |
|
|
8.7 |
Essentials |
|
|
8.8 |
References and Suggested Readings |
|
Part IV |
Applications |
|
9 |
General Comments on Chemical
Applications |
|
|
9.1 |
Introduction |
|
|
9.2 |
Classification |
|
|
9.3 |
Modeling |
|
|
9.4 |
Mapping |
|
|
9.5 |
Associations; Moving Window |
|
|
9.6 |
Overview of the Examples in Chapters 10 to
20 |
|
|
9.7 |
Essentials |
|
|
9.8 |
References and Suggested Readings |
|
|
10 |
Clustering of Multi-Component
Analytical Data for Olive Oils |
|
|
10.1 |
The Problem |
|
|
10.2 |
The Data |
|
|
10.3 |
Preliminary Exploration of Possible
Networks |
|
|
10.4 |
Learning to Make Predictions |
|
|
10.5 |
Concluding Remarks |
|
|
10.6 |
References and Suggested Readings |
|
|
11 |
The Reactivity of Chemical Bonds
and the Classification of Chemical Reactions |
|
|
11.1 |
The Problem and the Data |
|
|
11.2 |
Architecture of the Network for
Back-Propagation Learning |
|
|
11.3 |
Using an Experimental Design Technique to
Select the Training Set |
|
|
11.4 |
Application of the Kohonen Learning |
|
|
11.5 |
Application of the Trained Multilayer
Network |
|
|
11.6 |
Chemical Significance of the Kohonen Map |
|
|
11.7 |
Classification of Reactions: The Data |
|
|
11.8 |
Classification of Reactions: Results |
|
|
11.9 |
References and Suggested Readings |
|
|
12 |
HPLC Optimization of Wine
Analysis |
|
|
12.1 |
The Problem of Modeling |
|
|
12.2 |
Modeling the Mobile Phase for HPLC by a
Standard Method |
|
|
12.3 |
Modeling the Mobile Phase for HPLC by a Neural
Network |
|
|
12.4 |
Comparison of Networks with Identical
Architectures |
|
|
12.5 |
References and Suggested Readings |
|
|
13 |
Quantitative Structure-Activity
Relationships |
|
|
13.1 |
The Problem |
|
|
13.2 |
Dataset I |
|
|
13.3 |
Architecture and Learning Procedure |
|
|
13.4 |
Prospects of the Method |
|
|
13.5 |
Dataset II |
|
|
13.6 |
Structure Representation by Autocorrelation of
the Molecular Electrostatic Potential |
|
|
13.7 |
Verification of Structure Representation by
Unsupervised Learning |
|
|
13.8 |
Modeling of Biological Activity by Supervised
Learning |
|
|
13.9 |
Data Set III |
|
|
13.10 |
Structure Representation by Spectrum-Like
Uniform Representation |
|
|
13.11 |
Selection of the Most Important Variables Using
a Genetic Algorithm |
|
|
13.12 |
Cross-validation of the Counter-Propagation
Model Obtained by the Optimal Reduced
Representation |
|
|
13.13 |
References and Suggested Readings |
|
|
14 |
The Electrophilic Aromatic
Substitution Reaction |
|
|
14.1 |
The Problem |
|
|
14.2 |
The Data |
|
|
14.3 |
The Network |
|
|
14.4 |
Learning and Results |
|
|
14.5 |
A Third Representation of Data |
|
|
14.6 |
Concluding Remarks |
|
|
14.7 |
References and Suggested Readings |
|
|
15 |
Modeling and Optimizing a Recipe
for a Paint Coating |
|
|
15.1 |
The Problem |
|
|
15.2 |
The Data |
|
|
15.3 |
The Network and Training |
|
|
15.4 |
The Models |
|
|
15.5 |
References and Suggested Readings |
|
|
16 |
Fault Detection and Process
Control |
|
|
16.1 |
The Problems |
|
|
16.2 |
The Data |
|
|
16.3 |
The Methods |
|
|
16.4 |
Predictions of Faults |
|
|
16.5 |
Modeling and Controlling a Continuously Stirred
Tank Reactor (CSTR) |
|
|
16.6 |
References and Suggested Readings |
|
|
17 |
Secondary Structure of
Proteins |
|
|
17.1 |
The Problem |
|
|
17.2 |
Representation of Amino Acids as Input
Data |
|
|
17.3 |
Architecture of the Network |
|
|
17.4 |
Learning and Prediction |
|
|
17.5 |
References and Suggested Readings |
|
|
18 |
Infrared Spectrum-Structure
Correlation |
|
|
18.1 |
The Problem |
|
|
18.2 |
The Representation of Infrared Spectra as
Intensities |
|
|
18.3 |
The Dataset, and Learning by
Back-Propagation |
|
|
18.4 |
Adjustable Representation of an Infrared
Spectrum |
|
|
18.5 |
Representing Spectra using Truncated Sets of
Fourier or Hadamard Coefficients |
|
|
18.6 |
Results of Kohonen Learning |
|
|
18.7 |
A Molecular Transform of the 3D Structure |
|
|
18.8 |
Learning by Counter-Propagation |
|
|
18.9 |
Different Strategies for the Selection of a
Training Set |
|
|
18.10 |
From the Infrared Spectrum to the 3D
Structure |
|
|
18.11 |
References and Suggested Readings |
|
|
19 |
Properties of Molecular
Surfaces |
|
|
19.1 |
The Problems |
|
|
19.2 |
The Network Architecture and Training |
|
|
19.3 |
Tiling with Kohonen Maps; Conformational
Effects |
|
|
19.4 |
Investigation of Receptors of Biological Neural
Networks |
|
|
19.5 |
Comparison of Kohonen Maps |
|
|
19.6 |
Bioisosteric Design |
|
|
19.7 |
Molecular Shape Analysis |
|
|
19.8 |
References and Suggested Readings |
|
|
20 |
Libraries of Chemical
Compounds |
|
|
20.1 |
The Problems |
|
|
20.2 |
Strucure Coding |
|
|
20.3 |
Separation of Benzodiazepine and Dopamine
Agonists |
|
|
20.4 |
Finding Active Compounds in a Large Set of
Inactive Compounds |
|
|
20.5 |
Diversity and Similarity of Combinatorial
Libraries |
|
|
20.6 |
Deconvolution of Xanthene Sublibraries |
|
|
20.7 |
References and Suggested Readings |
|
|
21 |
Representation of Chemical
Structures |
|
|
21.1 |
The Problem |
|
|
21.2 |
Coding the Constitution |
|
|
21.3 |
Coding the 3D Structure I |
|
|
21.4 |
Coding the 3D Structure II |
|
|
21.5 |
Coding Molecular Surfaces |
|
|
21.6 |
A Hierarchy of Representations |
|
|
21.7 |
References and Suggested Readings |
|
22 |
Prospects of Neural Networks for
Chemical Applications |
|
|
Appendices |
|
|
* |
Programs |
|
|
* |
Data Sets |
|
|
* |
Presentation Material |
|
|
* |
Publications |
|
|
* |
Tutorials |