Counterpropagation

Figure 10.1-9. Architecture of the counterpropagation network for the classification of toxicants with one output layer.

Counterpropagation neural networks (CFG NN) were then used to establish relationships between protons and their H NMR chemical shifts. A detailed description of this method is given in the Tools Section 10,2.4.2,... 

A combination of physicochemical, topological, and geometric information is used to encode the environment of a proton, The geometric information is based on (local) proton radial distribution function (RDF) descriptors and characterizes the 3D environment of the proton. Counterpropagation neural networks established the relationship between protons and their h NMR chemical shifts (for details of neural networks, see Section 9,5). Four different types of protons were... 

These pairs of encoded structures and their (R spectra are used to ti ain a counterpropagation network (see Section 9.5.5). The two-layer netwoi k pi ocesses the structural information in its upper part and the spectral information in its lower part. Thus the network learns the correlation between the structures and their (R spec tra. This prnciedine is shown in Figine 10.2-8. 

Figure 10.2-S. Procedure for spectra simulation the query structure is coded, a training set of structure-spectra pairs is selected from the database, and the counterpropagation network is trained.

Figure 10.2-9. Application of a counterpropagation neural network as a look-up table for IR spectra sinnulation, The winning neuron which contains the RDF code in the upper layer of the network points to the simulated IR spectrum in the lower layer.

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. 

Several methods have been developed for establishing correlations between IR vibrational bands and substructure fragments. Counterpropagation neural networks were used to make predictions of the full spectra from RDF codes of the molecules. 

F. Ehientreich, M. Novic, S. Bohanec, J. Zupan, Bewertung von IR-Spek-trum-Struktur-Korrelationen mit Counterpropagation-Netzen, in Soft-ware-Entwicklung in der Chemie 10,... 

Several nonlinear QSAR methods have been proposed in recent years. Most of these methods are based on either ANN or machine learning techniques. Both back-propagation (BP-ANN) and counterpropagation (CP-ANN) neural networks [33] were used in these studies. Because optimization of many parameters is involved in these techniques, the speed of the analysis is relatively slow. More recently, Hirst reported a simple and fast nonlinear QSAR method in which the activity surface was generated from the activities of training set compounds based on some predefined mathematical functions [34]. 

J. Zupan, M. Novic and I. Ruisanchez, Kohonen and counterpropagation networks in analytical chemistry. Chemom. Intell. Lab. Syst., 38 (1997) 1-23. 

The system used to measure the optical fiber signals employs two separate frequency tunable laser light sources operating at about 1320 nm wavelength. One laser acts as a pump laser, whereas the other serves as the probe laser that sends optical pulses down the fiber to interact with the counterpropagating laser light wave pumped into the fiber from its opposite end. 

Another type of ANNs widely employed is represented by the Kohonen self organizing maps (SOMs), used for unsupervised exploratory analysis, and by the counterpropagation (CP) neural networks, used for nonlinear regression and classification (Marini, 2009). Also, these tools require a considerable number of objects to build reliable models and a severe validation. 

If v which means contribution from the absorption modulation is negligible, and the p-polarized probe beam counterpropagates along Beam 1, Eq. (8) can be simplified as... 

Two different experimental geometries are used for DFWM one with counterpropagating (Fig. 6a) and one with copropagating (Fig. 6b) laser pulses. The setup with counter propagating beams will be discussed first. 

The IR pulse is split into a weak probe beam, which passes down a computer-controlled variable delay line with up to 12 ns of delay and a strong pump beam. The pump and probe pulses are counterpropagating and focused into the center of the SCF cell. Typical spot sizes (1/e radius of E-field) were oj0 120 pm for the pump beam and oj0 60 pm for the probe beam. A few percent of the transmitted probe beam is split off and directed into an InSb detector. A reference beam is sent through a different portion of the sample. The reference beam is used to perform shot-to-shot normalization. The pump beam is chopped at half the laser repetition rate (900 Hz). The shot-to-shot normalized signal is measured with a lock-in amplifier and recorded by computer. 

Hecht-Nielsen, R. (1987). Counterpropagation networks. Applied Optics 26,4979-84. 

---