Neural Networks

Neural networks (NN) involves making decisions by learning the results of past performance. In that regard, NN is not much different from most of the methods discussed previously where experimentation is used to yield a data set from which, in turn, a predicting equation may be derived for future application. Although, as was pointed out earlier, instrumental approaches to relating sensory-to-instrumental data are often blind in the sense that the chemical/physical predictor(s) chosen may not be at all related to sensory quality, that risk is even greater for NN. NN consists of three parts: input (measurement) variables, hidden interrelations, and output variables (41,42). The output values are the result of examination of the thousands of permutations arising from even a moderate number of different measurement values. The process is spoken of as "training." By means of 10,000 or 100,000 iterations— or even more—interrelations among the hidden values are repeatedly evaluated until a set is found that correlates with the known differences in the test samples, Based on evaluation of numerous test samples, output values are eventually found that allow categorization of an unknown to possible classes, ie, the best product, a product matching the traditional product, one suffering from flaws such as off-flavor, or some one of the sensory notes being so dominant as to be objectionable. Once the necessary output value(s) have been obtained, NN may then be used for routine evaluation such as in quality control.

You Are What You Eat

You Are What You Eat

Nutrition is a matter that people spend their careers learning about and requires volumes of books to explain. My objective is to instruct you how to consume a healthy nutritional diet that aids your body in burning off fat instead of storing it. You do not require overwhelming science to get this.

Get My Free Ebook

Post a comment