G. Kateman

Multicomponent self-modelling curve resolution in high-performance liquid chromatography by iterative target transformation analysis

Abstract Iterative target transformation factor analysis can provide a method for resolving elution profiles consisting of any number of compounds. The results obtained for 3-component resolution are consistent with the results obtained with conventional methods of curve resolution. The same restrictions with regard to overlap and relative signal heights of the compounds seem to apply to the conventional method of curve resolution and the proposed method. The method is tested on data from high-performance liquid chromatography with a diode-array detector obtained for polynuclear aromatic hydrocarbons and for proteins.

Using artificial neural networks for solving chemical problems Part I. Multi-layer feed-forward networks

Abstract Smits, J.R.M., Melssen, W.J., Buydens, L.M.C. and Kateman, G., 1994. Using artificial neural networks for solving chemical problems. Part I. Multi-layer feed-forward networks. Chemometrics and Intelligent Laboratory Systems , 22: 165–189. This tutorial focuses on the practical issues concerning applications of different types of neural networks. The tutorial is divided into two parts. In the first part, an overview of the general appearance of neural networks is given and the multi-layer feed-forward neural network is described. In the second part, the Kohonen self-organising feature map and the Hopfield network are discussed. Since the multi-layer feed-forward neural network is one of the most popular networks, the theory concerning this network can easily be found in other references (B.J. Wythoff, Chemom. Intell. Lab. Syst. , 18 (1993) 115–155) and is therefore only described superficially in this paper. Much attention is paid to the practical issues concerning applications of the networks. For each network, a description is given of the types of problems which can be tackled by the specific neural network, followed by a protocol for the development of the system. It is seen that different neural networks are suited for different kinds of problems. Application of the networks is not always straightforward; a lot of constraints and conditions have to be fulfilled when using neural networks properly. They appear to be powerful techniques, but often a lot of experience is needed. In this paper some guidelines are given to avoid the most common difficulties in applying neural networks to chemical problems.

Multicriteria target vector optimization of analytical procedures using a genetic algorithm: Part I. Theory, numerical simulations and application to atomic emission spectroscopy

Abstract The target vector criterion was combined with a genetic algorithm as a parallel and evolutionary search technique. In this way a multicriteria optimization technique was obtained. In a number of numerical simulations the method was compared with classical optimization techniques used in analytical chemistry such as grid search, modified simplex, steepest ascent, overlapping resolution map, pattern search and simulated annealing according to their computational expenditure and speed, their ability to find global and local optima, their on-line applicability in analytical instrumentation and number of experiments needed. Then the technique was experimentally demonstrated for the simulataneous optimization of intensities of six atomic emission lines of trace elements in alumina powder as a function of spectroscopic excitation conditions.

Optimization of calibration data with the dynamic genetic algorithm

Abstract Genetic algorithms constitute a set of powerful search heuristics. A modified genetic algorithm was used to optimize calibration data sets. In order to construct an ideal genetic procedure, the diversity in a population is crucial. The idea proposed is to estimate the diversities along two directions, namely the diversity between the chromosomes in a population and the diversity between the alleles in all chromosomes. The newly defined diversity functions are able to describe the procedure of a genetic algorithm in detail and can be used as a feedback for dynamic control of the process in an almost ideal way. The optimization results show that for both short and long runs the dynamic genetic algorithm is superior to the “classical” genetic algorithms and that after optimization not only can the data sets be compacted and refined but also the predictive ability of the calibration model can be improved.

USING ARTIFICIAL NEURAL NETWORKS FOR SOLVING CHEMICAL PROBLEMS .2. KOHONEN SELF-ORGANIZING FEATURE MAPS AND HOPFIELD NETWORKS

This second part of a Tutorial on neural networks focuses on the Kohonen self-organising feature map and the Hopfield network. First a theoretical description of each type is given. The practical issues concerning applications of the networks are then discussed. For each network, a description is given of the types of problems which can be tackled by the specific neural network, followed by a protocol for the development of the neural network system. It is seen that different neural networks are suited for different kinds of problems. Guidelines to avoid common difficulties in using neural networks are also given.

Pattern classification with artificial neural networks : classification of algae, based upon flow cytometer data

Abstract In this paper the applicability of artificial neural networks as pattern classifiers is investigated. To study the behaviour of neural networks as pattern classifier for complex data, the identification and counting of phytoplankton, based upon flow cytometer data, is taken as an example. For this problem most conventional pattern recognition techniques fail, due to the shape of the clusters in the data. Three experiments have been carried out. First, it is investigated whether artificial neural systems are capable of discriminating between two different classes of algal species (poisonous species and non-poisonous species). Second, it is tested whether neural network systems can be used for identification of algal species. Third, the robustness of neural network systems towards changes in the flow cytometer settings has been studied. The results of these experiments show that a neural network system may be used as a pattern classifier.

Experimental optimization procedures in the determination of phosphate by flow-injection analysis

Abstract Flow injection analysis is attractive for handling large numbers of similar samples on a routine basis. In order to develop a procedure with a well defined performance in a relatively short time, application of experimental optimization procedures is useful. The two procedures evaluated are factorial design and simplex optimization. Criteria on the performance of a particular analytical method are formulated in terms of signal height, peak, width, baseline noise and linearity of the calibration grap. As a demonstration of the general approach, the determination of phosphate in aqueous solution is discussed; the flow rates of the water carrier stream and the reagent streams, the injection volume and the lengths of coils are the parameters applied for optimization of the procedure.

A Kalman filter for calibration, evaluation of unknown samples and quality control in drifting systems : Part 1. Theory and Simulations

Abstract The suitability of a Kalman filter for processing slowly varying parameters of a linear calibration graph is described. After calibration, the recursive algorithm predicts the changing parameters in time, which are used for the evaluation of unknown samples. With a preselected precision of the final results as a proper analytical goal, one may decide successively either to do another unknown sample or to calibrate again. The application of the proposed algorithm is demonstrated with a simulated example based on realistic data.

Two-dimensional mapping of IR spectra using a parallel implemented self-organising feature map

Abstract Melssen, W.J., Smits, J.R.M., Rolf, G.H. and Kateman, G., 1993. Two-dimensional mapping of IR spectra using a parallel implemented self-organising feature map. Chemometrics and Intelligent Laboratory Systems , 18: 195–204. A large data base containing 3284 infrared (IR) spectra (1327 wavelengths) of various molecules was investigated with a self-organising feature map (Kohonen network). In order to reduce the time required to train the network, a parallel implementation of the algorithm was developed. Application of the Kohonen network appears to be a powerful technique in mapping a high dimensional data space onto a two-dimensional one. Fragment coding was used to indicate the presence or absence of chemical functional groups in a molecule. Two-dimensional maps have been constructed for several fragments. Some preliminary results are presented in this paper. It appeared that some of the fragments were mapped onto relatively small regions (clusters) in the map. Mostly, these fragments were characterised by a high separability index, indicating that these functional groups were easily recognised by the network. Next, it was shown that, for some of the fragments which formed clusters in the map, a further differentiation into sub-fragments appeared to be possible. We conclude that the analysis of Kohonen maps yields valuable information which may be used for the practical design of a modular tree-like system of dedicated multi-layer feed-forward neural networks for the automated interpretation of infrared spectra.

Interpretation of infrared spectra with modular neural-network systems

Abstract The interpretation of infrared (IR) spectra is not straightforward and requires much time and expertise. In this study the interpretation of IR spectra using artificial neural networks is addressed. The conventional approach is to design a single neural network to cover the problem domain. In this study a different approach is taken by tackling specific sub-problems with small, dedicated neural networks. Such networks are intended to form the modules of a larger, structured system for spectrum interpretation. The problem domain chosen in this preliminary work concerned the decision on the presence or absence of various functional groups (i.e., alcohols and carbonyls). Several modules were created, as well as a large, ‘flat’ neural network which covered the entire problem domain. The performance of the specialized modules was compared with that of the flat network and with that of a genuine expert.