ART network

Neuronal networks are nowadays predominantly applied in classification tasks. Here, three kind of networks are tested First the backpropagation network is used, due to the fact that it is the most robust and common network. The other two networks which are considered within this study have special adapted architectures for classification tasks. The Learning Vector Quantization (LVQ) Network consists of a neuronal structure that represents the LVQ learning strategy. The Fuzzy Adaptive Resonance Theory (Fuzzy-ART) network is a sophisticated network with a very complex structure but a high performance on classification tasks. Overviews on this extensive subject are given in [2] and [6]. 

Many different types of networks have been developed. They all consist of small units, neurons, that are interconnected. The local behaviour of these units determines the overall behaviour of the network. The most common is the multi-layer-feed-forward network (MLF). Recently, other networks such as the Kohonen, radial basis function and ART networks have raised interest in the chemical application area. In this chapter we focus on the MLF networks. The principle of some of the other networks are explained and we also discuss how these networks relate with other algorithms, described elsewhere in this book. 

ART networks consist of units that contain a weight vector of the same dimension as the input patterns. Each unit is meant to represent one class or cluster in the input patterns. The structure of the ART network is such that the number of units is larger than the expected number of classes. The units in excess are dummy units that can be taken into use when a new input pattern shows up that does not belong to any of the already learned classes. 

The structure of ART networks is hard to visualize. It is in fact a theory that can better be explained by means of a sequence of steps that follow the strategy. 

From the previous discussion, it is clear that ART networks are more applicable for pattern recognition than for quantitative applications. In Fig. 44.27 it is shown how ART networks functions with different values for p. It has been applied to... 

Fig. 44.27. The influence of different values of p on the classification performance of the ART network. (a) large p value (b) small p values.

Procedurally, an input data vector is presented to the ART network. 

No single ART2 equation or set of equations provides the desired adaptive properties for the system. Rather, it is the synergistic effect of this architecture as a whole that gives rise to these properties. A description of the ART network applied to process situations can be found in Whiteley and Davis (1996) a complete description of an ART network can be found in Carpenter and Grossberg (1987a). 

Most uses of ART networks by the chemical community are very recent. In applications related to infrared spectroscopy, they have been used to recognize aromatic substitution pattems (they reportedly performed better than human experts) and also in the clustering of spectra of lubricating base oils. Reference 83 also contains a good comparison of ART networks with other methods. ART networks have also been applied in choosing a detector for ion... 

BAM networks are two-layer feedforward/feedback heteroassodative networks (they can also be autoassociative). An example network is shown in Figure 6. Standard BAMs take bipolar ( 1) inputs and outputs. Adaptive BAMs (ABAM) can take continuous inputs and outputs. In either case, input data should be mutually orthogonal (i.e., independent, nonredundant, and uncorrelated). BAMs were inspired by ART networks but are conceptually sim-... 

Domine and co-workers utilized the family of Adaptive Resonance Theory (ART and ART 2-A) based artificial neural networks for unsupervised and supervised pattern recognition (142,143). The simplest ART network is a vec-... 

Domine, D., Devillers, J., Wienke, D., and Buydens, L. (1996) The heuristic potency of art networks for QSAR data visualization and interpretation. Abstr. Pap. Am. Chem. Soc. 211th, COMP-108. 

More than 50 different types of neural network exist. Certain networks are more efficient in optimization others perform better in data modeling and so forth. According to Basheer (2000) the most popular neural networks today are the Hopfield networks, the Adaptive Resonance Theory (ART) networks, the Kohonen networks, the counter propagation networks, the Radial Basis Function (RBF) networks, the backpropagation networks and recurrent networks. 

---