Pattern Recognition in Chemistry

ISBN-13: 9783540102731 / Angielski / Miękka / 1980 / 222 str.

Kurt Varmuza

Opis

Analytical chemistry of the recent years is strongly influenced by automation. Data acquisition from analytica instruments - and some- times also controlling of instruments - by a computer are principally solved since many years. Availability of microcomputers made these tasks also feasible from the economic point of view. Besides these basic applications of computers in chemical measurements scientists developed computer programs for solving more sophisticated problems for which some kind of "intelligence" is usually supposed to be necessary. Harm- less numerical experiments on this topic led to passionate discussions about the theme "which jobs cannot be done by a computer but only by human brain ? . If this question is useful at all it should not be ans- wered a priori. Application of computers in chemistry is a matter of utility, sometimes it is a social problem, but it is never a question of piety for the human brain. Automated instruments and the necessity to work on complex pro- blems enhanced the development of automatic methods for the reduction and interpretation of large data sets. Numerous methods from mathematics, statistics, information theory, and computer science have been exten- sively investigated for the elucidation of chemical information; a new discipline "chemometrics" has been established. Three different approaches have been used for computer-assisted interpretations of chemical data. 1. Heuristic methods try to formu- late computer programs working in a similar way as a chemist would solve the problem. 2.

Spis treści

A: Introduction to Some Pattern Recognition Methods.- 1. Basic Concepts.- 1.1. First Ideas of Pattern Recognition.- 1.2. Pattern Space.- 1.3. Binary Classifiers.- 1.4. Training and Evaluation of Classifiers.- 1.5. Additional Aspects.- 1.6. Warning.- 1.7. Applications of Pattern Recognition.- 1.8. Literature.- 1.8.1. General Pattern Recognition.- 1.8.2. Pattern Recognition from the Chemist’s Point of View.- 2. Computation of Binary Classifiers.- 2.1. Classification by Distance Measurements to Centres of Gravity.- 2.1.1. Principle.- 2.1.2. Centres of Gravity in a d-DimensionaI Space.- 2.1.3. Classification by Distance Measurements.- 2.1.4. Classification by the Symmetry Plane.- 2.1.5. Classification by Mean Vectors.- 2.1.6. Evaluation.- 2.1.7. Projection of Pattern Points on a Hypersphere.- 2.1.8. Distance Measurements in Pattern Space (Overview),.- 2.1.9. Distance Measurements with Weighted Features (Generalized Distances).- 2.1.10. Chemical Applications.- 2.2. Learning Machine.- 2.2.1. Principle.- 2.2.2. Initial Weight Vector.- 2.2.3. Correction of the Weight Vector.- 2.2.4. Methods of Training.- 2.2.5. Restrictions.- 2.2.6. Evaluation.- 2.2.7. Dead Zone Training.- 2.2.8. Chemical Applications.- 2.3. Linear Regression (Least-Squares Classification).- 2.3.1. Principle.- 2.3.2. Mathematical Treatment.- 2.3.3. Characteristics and Variations of the Method.- 2.3.4. Chemical Applications.- 2.4. Simplex Optimization of Classifiers.- 2.4.1. Principle.- 2.4.2. Starting the Simplex.- 2.4.3. Response Function.- 2.4.4. Moving the Simplex.- 2.4.5. Halting the Simplex.- 2.4.6. Chemical Applications.- 2.5. Piecewise-Linear Classifiers.- 2.6. Implementation of Binary Classifiers.- 2.6.1. Discrete or Continuous Response.- 2.6.2. Classification by a Committee of Classifiers.- 2.6.3. MuIticategory Classification.- 3. K — Nearest Neighbour Classification (KNN-Method).- 3.1. Principle.- 3.2. Maximum Risk of KNN-Classifications.- 3.3. Characteristics and Variations of the KNN-Method.- 3.4. Classification with Potential Functions.- 3.5. KNN-Classification with a Condensed Data Set.- 3.6. Chemical Applications.- 4. Classification by Adaptive Networks.- 4.1. Perceptron.- 4.2. Adaptive Digital Learning Network.- 4.3. Chemical Applications.- 5. Parametric Classification Methods.- 5.1. Principle.- 5.2. Bayes- and Maximum Likelihood Classifiers.- 5.3. Estimation of Probability Densities.- 5.4. Bayes- and Maximum Likelihood Classifiers for Binary Encoded Patterns.- 5.5. A Simple Sequential Classification Method Based on Probability Densities.- 5.6. Chemical Applications.- 6. Modelling of Clusters.- 6.1. Principle.- 6.2. Modelling by a Hypersphere.- 6.3. SIMCA-Method.- 7. Clustering Methods.- 7.1. Principle.- 7.2. Hierarchical Clustering.- 7.3. Minimal Spanning Tree Clustering.- 7.4. Chemical Applications.- 8. Display Methods.- 8.1. Principle.- 8.2. Linear Methods.- 8.3. Nonlinear Methods.- 8.4. Chemical Applications.- 9. Preprocessing.- 9.1. Principle.- 9.2. Scaling.- 9.3. Weighting.- 9.4. Transformation.- 9.5. Combination of Features.- 10. Feature Selection.- 10.1. Principle.- 10.2. Feature Selection by Using Data Statistics.- 10.2.1. Variance Weighting.- 10.2.2. Fisher Weighting.- 10.2.3. Use of Probability Density Curves.- 10.2.4. Methods for Binary Encoded Patterns.- 10.2.5. Elimination of Correlations.- 10.3. Feature Selection by Using Classification Results.- 10.3.1. Evaluation of the Components of a Single Weight Vector.- 10.3.2. Feature Selection with the Learning Machine.- 10.3.3. Other Methods.- 10.4. Number of Intrinsic Dimensions and Number of Patterns (n/d — Problem).- 11. Evaluation of Classifiers.- 11.1. Principle.- 11.2. Predictive Abilities.- 11.3. Loss.- 11.4. A posteriori Probabilities.- 11.5. Terminology Problems.- 11.6. Application of Information Theory.- 11.6.1. Introduction to Information Theory.- 11.6.2. Transinformation.- 11.6.3. Figure of Merit.- 11.7. Evaluation of Classifiers with Continuous Response.- 11.8. Confidence of Predictive Abilities.- 11.9. Comparison with the Capability of Chemists.- B: Application of Pattern Recognition Methods in Chemistry.- 12. General Aspects of Pattern Recognition in Chemistry.- 13. Spectral Analysis.- 13.1. Mass Spectrometry.- 13.1.1. Survey.- 13.1.2. Representation of Mass Spectra as Pattern Vectors.- 13.1.3. Determination of Molecular Formulas and Molecular Weights.- 13.1.4. Recognition of Molecular Structures.- 13.1.5. Chemical Interpretation of Mass Spectral Classifiers.- 13.1.6. Simulation of Mass Spectra.- 13.1.7. Miscellaneous.- 13.2. Infrared Spectroscopy.- 13.3. Raman Spectroscopy.- 13.4. Nuclear Magnetic Resonance Spectroscopy.- 13.5. Gamma-Ray Spectroscopy.- 13.6. Combined Spectral Data.- 14. Chromatography.- 14.1. Gas Chromatography.- 14.2. Thin Layer Chromatography.- 15. Electrochemistry.- 16. Classification of Materials and Chemical Compounds.- 16.1. Technology.- 16.2. Archaeology.- 16.3. Food.- 16.4. Biology.- 16.5. Chemistry.- 17. Relationships between Chemical Structure and Biological Activity.- 17.1. Pharmacological Activity.- 17.1.1. General Remarks.- 17.1.2. Sedatives and Tranquilizers.- 17.1.3. Cancer and Tumors.- 17.1.4. Miscellaneous.- 17.2. Odour Classification.- 17.3. Spectra — Activity Relationships.- 18. Clinical Chemistry.- 19. Environmental Chemistry.- 19.1. Petroleum Pollutants.- 19.2. Atmospheric Particulates.- 19.3. Miscellaneous.- 20. Classification of Analytical Methods.- C: Append.- 21. Literature.- 21.1. Pattern Recognition in Chemistry.- 21.2. General Pattern Recognition.- 21.3. Other Literature.- 21.4. List of Authors.- 22. Subject Index.