Algorithms for Chemists

This text has been designed to introduce chemists to various computer methods. It aims to give chemists an overview of the basic techniques, methods and algorithms used for handling and understanding information on a computer. Emphasis is given to methods dealing with sets of data rather than with individual objects. The book is divided into three parts: data, preprocessing of data and data handling. In the first part, the representation of data in multi-dimensional measurement spaces, space metrics, the ways in which data can be stored in files and how those files can be manipulated are discussed. The second part contains a description of elementary methods of "polishing" data before "real" feature extraction is attempted. These methods include smoothing, differentiation, peak detection, integration and on-line correlation. Basic procedures for preprocessing are introduced, applicable in a variety of situations, and methods such as autocorrelation and principal component analysis are discussed. Emphasis is given to the Fast Fourier Transform (FFT) algorithm because of its wide applicability, but most of the methods described are suited for analyzing a large number of complex data.
The third part of the book describes data handling, focusing on methods for clustering, pattern recognition, feature and knowledge extraction, learning and prediction of properties and expert systems. Ideas for the programming of fragment codes, connection tables and linear notations are described, together with an overview of fractal geometry and the fractal process. The section on expert systems concludes with an example of how most of the methods explained in the book can be used in the design of a chemical expert system based on different spectroscopies. The algorithms described are written as a mixture of BASIC, FORTRAN, PASCAL and Knuth's MIX, and should be understood by anyone with a basic knowledge of computer programming.