Soft modeling techniques

A.P. de Weyer, L.M.C. Buydens, G. Kateman and H.M. Heuvel, Neural networks used as a soft modelling technique for quantitative description of the inner relation between physical properties and mechanical properties of poly ethylene terephthalate yams. Chemom. Intell. Lab. Syst., 16(1992) 77-82. 

Various attempts have been made to use pattern recognition [24, 25] in QSAR studies and successful applications have been reported. Soft modeling techniques, e.g. the partial least squares (PLS) method [26, 27], now offer better opportunities. With the help of this principal component-like method the explanatory power of many, even hundreds or thousands of variables can be used for a limited number of objects, a task being absolutely impossible in regression analysis in which the number of objects must always be larger than the number of variables. 

In the A-matrix approach, all absorbing constituents of a sample must be explicitly known to be included into the calibration procedure. As we will see in the following, with more soft modeling techniques, it will also be possible to account for unknown constituents without their explicit calibration. 

Visual inspection should be possible from plots of predicted versus measured concentrations, from principal component plots of loadings and scores in the case of soft modeling techniques, and by plotting the standard error of calibration (SEC) or the standard error of prediction (SEP(-y, Eq. (6.68)) from cross-validation in dependence on the number of eigenvalues or of principal components. 

A review was described a new generation of contrast agents, which could be developed by making greater use of soft modeling techniques such as... 

Because the application of a neural network as a soft modeling technique for prediction purposes is rather new [6], in the following section, a short introduction to neural networks is given. If the reader is not interested in this aspect, this subject can be skipped without encountering problems when reading the remaining part of this chapter. 

Mass spectrometry methods based on soft ionization techniques, 59,61,88,89 matrix-assisted laser desorption ionization/time-of-flight (MALDI-TOF), have been successfully applied for the direct analysis of grape and wine extracts and for monitoring flavonoid reactions in model solution studies. They give access to the molecular weights of the different species present in a fraction or extract and, through fragmentation patterns, provide important information on their constitutive units. Description of the various MS techniques can be found in Chapters 1 and 2. 

Nevertheless, in most of the electronic tongue applications found in the literature, classification techniques like linear discriminant analysis (LDA) and partial least squares discriminant analysis (PLS-DA) have been used in place of more appropriate class-modeling methods. Moreover, in the few cases in which a class-modeling technique such as soft independent modeling of class analogy (SIMCA) is applied, attention is frequently focused only on its classification performance (e.g., correct classification rate). Use of such a restricted focus considerably underutilizes the significant characteristics of the class-modeling approach. 

Model-based nonlinear least-squares fitting is not the only method for the analysis of multiwavelength kinetics. Such data sets can be analyzed by so-called model-free or soft-modeling methods. These methods do not rely on a chemical model, but only on simple physical restrictions such as positiveness for concentrations and molar absorptivities. Soft-modeling methods are discussed in detail in Chapter 11 of this book. They can be a powerful alternative to hard-modeling methods described in this chapter. In particular, this is the case where there is no functional relationship that can describe the data quantitatively. These methods can also be invaluable aids in the development of the correct kinetic model that should be used to analyze the data by hard-modeling techniques. 

Esteban, M., Anno, C., Dfaz-Cruz, J.M., Dfaz-Cruz, M.S., and Tauler, R., Multivariate curve resolution with alternating least squares optimization a soft-modeling approach to metal complexation studies by voltammetric techniques, Trends Anal. Chem., 19, 49-61, 2000. 

Stored in a table where columns are descriptors, and rows are compounds (or conformers), QSAR data sets contain separate columns for the measured target property (Y), attributed to the training set, as well as computed descriptors for (external) reference compounds on which the QSAR model is tested—the test set. Statistical procedures, e.g., multiple linear regression (MLR), projection to latent structures (PLS), or neural networks (NN) [38], are then used to establish a mathematical soft model relating the observed measurement(s) in the Y column(s) with some combination of the properties represented in the subsequent columns. PLS, NN, and AI (artificial intelligence) techniques have been explored by Green and Marshall in the context of 3D-QSAR models [39], and were shown to extract similar information. A problem that may lead to spurious (chance) correlations when using MLR techniques, the colinearity between various descriptors, or cross-correlation, is usually dealt with in PLS [40],... 

The previous section dealt with application of soft computing techniques for developing predicting models for drilling forces and drilling-induced damage in PMCs. This section... 

---