Neural network structure

Fig. 10.43 Neural network structure for Example 10.9. Hidden layer...

The selection of cluster number, which is generally not known beforehand, represents the primary performance criterion. Optimization of performance therefore requires trial-and-error adjustment of the number of clusters. Once the cluster number is established, the neural network structure is used as a way to determine the linear discriminant for interpretation. In effect, the RBFN makes use of known transformed features space defined in terms of prototypes of similar patterns as a result of applying /c-means clustering. 

L. Luo, Predictability comparison of four neural network structures for correcting matrix effects in X-ray fluorescence spectrometry, J. Trace Microprobe Tech., 18(3), 2000, 349-360. 

In theory one hidden layer neural network is sufficient to describe all input/output relations. More hidden layers can be introduced to reduce the number of neurons compared to the number of neurons in a single layer neural network. The same argument holds for the type of activation function and the choice of the optimisation algorithm. However, the emphasis of this work is not directed on the selection of the best neural network structure, activation function and training protocol, but to the application of neural networks as a means of non-linear function fit. 

Rossen, K., Thomsen, H.K., Gniadecki, R., Hansen, LK. andWulf H.C. (2004) Melanoma diagnosis by Raman spectroscopy and neural networks structure alteration in proteins and lipids in intact cancer tissue. J. Invest. Dermatol, 122, 443-9. 

Feedback error learning (FEL) is a hybrid technique [113] using the mapping to replace the estimation of parameters within the feedback loop in a closed-loop control scheme. FEL is a feed-forward neural network structure, under training, learning the inverse dynamics of the controlled object. This method is based on contemporary physiological studies of the human cortex [114], and is shown in Figure 15.6. 

According to the relationship structure of basic safety evaluation factors of oil depot, hierarchy structure neural network of safety evaluation is built and is composed of 2 layers and 4 neural network units, the parameters of neural network structure is shown in table 1. 

FIGURE 3.9 Typical artificial neural network structure. 

Beside trial and error methods, other methodologies were tested in order to obtain the best performance of the neural models evolutionary algorithms represent appropriate methods for determining optimal neural network structure. 

Artificial neural network Group method of data handling Least-squares technique Liquid-liquid equilibrium Mass fractions Neural network structure Quaternary system... 

R. B. Boozarjomehry and W. Y. Svrcek, Automatic design of neural network structures, Comput. Chem. Engin., 2001, 25, 1075-1088. 

FIGURE 6.8 A neural network structure for a multiple input and single output problems. 

The results are shown for different neural network structures. The number of neurons the input and output layer is always the same, respectively 4 and 1, and the change in the number of hidden layers and the number of them occurring in neurons. 

W. Kaminski, P. Strumiilo and E. Tomczak, Optimization of Neural Network Structure in Drying Process Modelling, Proc. 2" Conf Neural Net Appl., Szczyrk 96, 1, (1996) 248-254. 

We have presented a neural network based spectrum classifier (NSC) aimed at ultrasonic resonance spectroscopy. The ultrasonic spectroscopy and the NSC has been evaluated in many industrial applications, such as concrete inspection, testing of aerospace composite structures, ball bearings, and aircraft multi-layer structures. The latter application has been presented in some detail. 

A challenging task in material science as well as in pharmaceutical research is to custom tailor a compound s properties. George S. Hammond stated that the most fundamental and lasting objective of synthesis is not production of new compounds, but production of properties (Norris Award Lecture, 1968). The molecular structure of an organic or inorganic compound determines its properties. Nevertheless, methods for the direct prediction of a compound s properties based on its molecular structure are usually not available (Figure 8-1). Therefore, the establishment of Quantitative Structure-Property Relationships (QSPRs) and Quantitative Structure-Activity Relationships (QSARs) uses an indirect approach in order to tackle this problem. In the first step, numerical descriptors encoding information about the molecular structure are calculated for a set of compounds. Secondly, statistical and artificial neural network models are used to predict the property or activity of interest based on these descriptors or a suitable subset. 

A structure descriptor is a mathematical representation of a molecule resulting from a procedure transforming the structural information encoded within a symbolic representation of a molecule. This mathematical representation has to be invariant to the molecule s size and number of atoms, to allow model building with statistical methods and artificial neural networks. 

The data analysis module of ELECTRAS is twofold. One part was designed for general statistical data analysis of numerical data. The second part offers a module For analyzing chemical data. The difference between the two modules is that the module for mere statistics applies the stati.stical methods or rieural networks directly to the input data while the module for chemical data analysis also contains methods for the calculation ol descriptors for chemical structures (cl. Chapter 8) Descriptors, and thus structure codes, are calculated for the input structures and then the statistical methods and neural networks can be applied to the codes. 

The possibilities for the application for neural networks in chemistry arc huge [10. They can be used for various tasks for the classification of structures or reactions, for establishing spcctra-strncturc correlations, for modeling and predicting biological activities, or to map the electrostatic potential on molecular surfaces. 

As explained in Chapter 8, descriptors are used to represent a chemical structure and, thus, to provide a coding which allows electronic processing of chemical data. The example given here shows how a GA is used to Rnd an optimal set of descriptors for the task of classification using a Kohoncii neural network. The chromosomes of the GA are to be used as a means for selecting the descriptors they indicate which descriptors are used and which are rejected ... 

The GA was then applied to select those descriptors which give the best classification of the structures when a Kohonen network is used. The objeetive function was based on the quality of the classification done by a neural network for the I educed descriptors. 

The same structure representation as the one taken in the original study [39] is selected in order to show some possibilities evolving from working with a neural network method. Tabic 10.1-1 gives the ten descriptors chosen lor the representation of the 115 molecules of the data set. 

Neural networks can learn automatically from a data set of examples. In the case of NMR chemical shiffs, neural networks have been trained to predict the chemical shift of protons on submission of a chemical structure. Two main issues play decisive roles how a proton is represented, and which examples are in the data set. 

A hands-on experience with the method is possible via the SPINUS web service [48. This service uses a client-server model. The user can draw a molecular structure within the web browser workspace (the client), and send it to a server where the predictions are computed by neural networks. The results are then sent back to the user in a few seconds and visualised with the same web browser. Several operations and different types of technology arc involved in the system ... 

Other methods consist of algorithms based on multivariate classification techniques or neural networks they are constructed for automatic recognition of structural properties from spectral data, or for simulation of spectra from structural properties [83]. Multivariate data analysis for spectrum interpretation is based on the characterization of spectra by a set of spectral features. A spectrum can be considered as a point in a multidimensional space with the coordinates defined by spectral features. Exploratory data analysis and cluster analysis are used to investigate the multidimensional space and to evaluate rules to distinguish structure classes. 

Neural networks have been applied to IR spectrum interpreting systems in many variations and applications. Anand [108] introduced a neural network approach to analyze the presence of amino acids in protein molecules with a reliability of nearly 90%. Robb and Munk [109] used a linear neural network model for interpreting IR spectra for routine analysis purposes, with a similar performance. Ehrentreich et al. [110] used a counterpropagation network based on a strategy of Novic and Zupan [111] to model the correlation of structures and IR spectra. Penchev and co-workers [112] compared three types of spectral features derived from IR peak tables for their ability to be used in automatic classification of IR spectra. 

In many cases, structure elucidation with artificial neural networks is limited to backpropagation networks [113] and, is therefore performed in a supervised man-... 

---