Pattern recognition methods results

An invariant pattern recognition method, based on the Hartley transform, and applied to radiographic images, containing different types of weld defects, is presented. Practical results show that this method is capable to describe weld flaws into a small feature vectors, allowing their recognition automatically by the inspection system we are realizing. 

Yeh and Spiegelman [24], Very good results were also obtained by using simple neural networks of the type described in Section 33.2.9 to derive a decision rule at each branching of the tree [25]. Classification trees have been used relatively rarely in chemometrics, but it seems that in general [26] their performance is comparable to that of the best pattern recognition methods. 

Advanced mathematical and statistical techniques used in analytical chemistry are often referred to under the umbrella term of chemometrics. This is a loose definition, and chemometrics are not readily distinguished from the more rudimentary techniques discussed in the earlier parts of this chapter, except in terms of sophistication. The techniques are applied to the development and assessment of analytical methods as well as to the assessment and interpretation of results. Once the province of the mathematician, the computational powers of the personal computer now make such techniques routinely accessible to analysts. Hence, although it would be inappropriate to consider the detail of the methods in a book at this level, it is nevertheless important to introduce some of the salient features to give an indication of their value. Two important applications in analytical chemistry are in method optimization and pattern recognition of results. 

Next, supervised-learning pattern recognition methods were applied to the data set. The 111 bonds from these 28 molecules were classified as either breakable (36) or non-breakable (75), and a stepwise discriminant analysis showed that three variables, out of the six mentioned above, were particularly significant resonance effect, R, bond polarity, Qa, and bond dissociation energy, BDE. With these three variables 97.3% of the non-breakable bonds, and 86.1% of the breakable bonds could be correctly classified. This says that chemical reactivity as given by the ease of heterolysis of a bond is well defined in the space determined by just those three parameters. The same conclusion can be drawn from the results of a K-nearest neighbor analysis with k assuming any value between one and ten, 87 to 92% of the bonds could be correctly classified. 

A generalised structure of an electronic nose is shown in Fig. 15.9. The sensor array may be QMB, conducting polymer, MOS or MS-based sensors. The data generated by each sensor are processed by a pattern-recognition algorithm and the results are then analysed. The ability to characterise complex mixtures without the need to identify and quantify individual components is one of the main advantages of such an approach. The pattern-recognition methods maybe divided into non-supervised (e.g. principal component analysis, PCA) and supervised (artificial neural network, ANN) methods also a combination of both can be used. 

Reference spectra choice is critical when applying supervised pattern recognition methods. The first solution is to use pure compound spectra as references. The drawback is that mixture spectra in data cubes often differ from the reference spectra. Applying the model may therefore give wrong results. The second solution, suitable in a few studies, is to select image pixels where only one compound is present in order to obtain the calibration sets. 

In order to characterise wine samples into the mentioned four classes, a supervised pattern recognition method (LDA) was applied. The results obtained gave 100% correct classification for the three classes (Barbera Oltrepo, Barbera Piemonte and Barbera Alba) and only one Barbera Asti sample was not correctly classified (cross-validation error rate 1.89%). 

Using the results in Table XII the proposed pattern recognition method was applied in discriminating Florida from Brazil... 

Electronic noses provide new possibilities for monitor the state of a cultivation non-in-vasively in real-time. The electronic nose uses an array of chemical gas sensors that monitors the off-gas from the bioreactor. By taking advantage of the off-gas components different affinities towards the sensors in the array it is possible with the help of pattern recognition methods to extract valuable information from the culture in a way similar to the human nose. For example, with artificial neural networks, metabolite and biomass concentration can be predicted, the fermentability of a medium before starting the fermentation estimated, and the growth and production stages of the culture visualized. In this review these and other recent results with electronic noses from monitoring microbial and cell cultures in bioreactors are described. 

HTS produces large numbers of individual measurements with inherent variability and error reflected in confirmation rates often significantly below 100%. Statistical and pattern recognition methods to analyze HTS data and optimize assay parameters are routinely used. (Padmanabha, Cook, and Gill, 2005). Various error sources can influence the variability of HTS data, leading to false positive and false negative results (Parker and Bajorath, 2006 Makarenkov et al., 2007). 

The aim of supervised classification is to create rules based on a set of training samples belonging to a priori known classes. Then the resulting rules are used to classify new samples in none, one, or several of the classes. Supervised pattern recognition methods can be classified as parametric or nonparametric and linear or nonlinear. The term parametric means that the method makes an assumption about the distribution of the data, for instance, a Gaussian distribution. Frequently used parametric methods are EDA, QDA, PLSDA, and SIMCA. On the contrary, kNN and CART make no assumption about the distribution of the data, so these procedures are considered as nonparametric. Another distinction between the classification techniques concerns the... 

The choice of properties has a major influence on pattern recognition methods and different property sets will result in different patterns of similarity between compounds. Several methods are available to make selections of subsets of uncorrelated properties which can be used for QSAR studies. These include ... 

We haveemployed a variety of unsupervised and supervised pattern recognition methods such as principal component analysis, cluster analysis, k-nearest neighbour method, linear discriminant analysis, and logistic regression analysis, to study such reactivity spaces. We have published a more detailed description of these investigations. As a result of this, functions could be developed that use the values of the chemical effects calculated by the methods mentioned in this paper. These functions allow the calculation of the reactivity of each individual bond of a molecule. 

Classification by distance measurements to centres of gravity requires compact clusters which are often not present in chemical applications. Usually, other classification methods give better results. Nevertheless, this simple and evident method serves as a standard for comparisons with more sophisticated pattern recognition methods. 

Many objections against the usefulness of pattern recognition methods for chemical problems are legitimated because the statistical evaluation was performed unsatisfactorily in many papers. Further confusions result from a non-uniform terminology (Chapter 11.5). An objective mathematical evaluation of classifiers is an absolute necessary prerequisite to a further application to actual classification problems. 

Crawford and Morrison C603 found for a sophisticated mass spectral interpretation program a capability of the same order as that for an undergraduate student- A similar result has been reported C1193 about the interpretation of binary encoded infrared spectra. Kowalski et. al. C162 l emphasized the superiority of pattern recognition methods in the interpretation of multidimensional data. 

Additional application of chemical knowledge to the selection of features or to the classifier construction has improved the classification results C1933. A comparison between pattern recognition methods and a sophisticated interpretative library search system for mass spectra ( STIRS C39, 4221) has indicated some superiority of the STIRS-system C172, 202, 3321. A decision tree pattern recognition was recommended by Neisel et. al. C2051 as a supplement to library search. 

These examples show that improper use of pattern recognition methods makes it possible to obtain good classification results even for senseless questions. Normally, such effects occur if the number of spectra is a) too small, b) too different in the classes or c) does not exceed significantly the number of dimensions C251]. 

---