Customized neural networks

Current trends in neural networks favor smaller networks with minimal architecture. Two major advantages of smaller networks previously discussed are better generalization capability (8.4) and easier rule extraction (13.2). Another advantage is better predictive accuracy, seen when a large network is replaced by many smaller networks, each for a subtask or a subset of data. A typical example is the protein classification problem, where n individual networks can be used to classify n different protein families and increase the prediction accuracy obtained by one large network with n output units. The improvement is especially significant when there is sufficient data for fine-tuning individual neural networks to the particularity of the data subsets. The use of ensembles of small, customized neural networks to improve predictive accuracy has been shown in numerous cases. 

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. 

Figure 15-7 Neural networks provide ultimate freedom of discrimination, with the additional danger of overfitting a solution.The curved line separating the two populations for this specific set of data may not work well in a subsequent sample since it is custom molded to separate this test set.

One commercially available sensor array analysis system, offered by Mosaic Industries [51], is Rhino , a microprocessor-based instrument with an array composed of discrete, resistive gas sensors. An artificial neural network processes sensor inputs and relates them to patterns established by training the instrument with gas components and mixtures of interest for a specific application. In principle, each system is customized for an application by the choice of sensors and the gas detection needs. Potential applications for this system are limited by the availability of suitable sensors and the complexity needed for discrimination. 

Mass Customization 836 Neural Network 911 Optimization in Manufacturing 929 Optimization in Manufacturing Systems, Fundamentals 932 Planning 946 Process 973 Production 995 Production Planning 1003 Productivity 1006 Reconfigurable Manufacturing System... 

An Assessment of Customer Contentment for Ready-to-Drink Tea Flavor Notes Using Artificial Neural Networks... 

The objective of this study, therefore, is to assess which compound, which affects customer contentment for RTD tea flavor notes, would be successful in the market. In order to identify the hidden pattern of the customer s needs, the artificial neural networks (ANNs) have been applied to classify the key volatile compound of the flavors. In this study, the RTD tea segment is selected as a case study due to market trend. The 4 customer groups and 5 compounds of flavor are used as a data. 

Data obtained in such a way for over 400 catalyst formulations has initially been used to train artificial neural networks, which in turn allow the prediction of catalyst performance of yet unprepared catalyst compositions. This approach is nowadays a standard tool when catalysts have to be designed for the special process conditions of a given customer. This broad database combined with artificial neural networks allows Clariant to customize its PHTHALIMAX catalyst for phthalic anhydride production for each customer process. 

Fault tolerance. If the architecture of the network is damaged, the network will retain some functionality. This is more important if the neural network is implemented in custom hardware rather than on a desktop computer. 

The system in Fig. 61 is typical of a hybrid opto-electronic neural network [71], It could equally well be implemented with optical fibers or other guided wave optics. A recent experimental neural network based on Fig. 61 was constructed with a custom FLC SLM. As the system uses one channel from each neuron, the number of pixels required for the weights is often small. In this case the SLM was an 8x8 array of 5 mm pixels. The photodetector was made from an array of amorphous silicon photoresistors, but these proved too nonuniform and so were replaced with an array of commercial-... 

Neural Nets (NNs) relate a set of input neurons with an output neuron (providing the prediction label of a data point) by a network of layers of neurons in the interior. They are certainly among the most frequently used Machine Learning methods in the field [148] and allow for a high degree of customization since the architecture of the network itself is part of the parameters the user may define. 

---