What Are Artificial Neural Networks

Artificial neural networks are now widely used in science. Not only are they able to learn by inspection of data rather than having to be told what to do, but they can construct a suitable relationship between input data and the target responses without any need for a theoretical model with which to work. For example, they are able to assess absorption spectra without knowing about the underlying line shape of a spectral feature, unlike many conventional methods. 

J. Zupan, Introduction to artificial neural network (ANN) methods what they are and how to use them, Acta Chim. Slav., 41(3), 1994, 327-352. 

Many instruments aheady have some chemometrics routines built into their software in such a way that their use is totally transparent to the final user (and sometimes the word chemometrics is not even mentioned, to avoid possible aversion). Of course, they are closed routines, and therefore the user cannot modify them. It is quite obvious that it would be much better if chemometric knowledge were much more widespread, in order that the user could better understand what kind of treatment the data have undergone and eventually modify the routines in order to make them more suitable to the user s requirements. As computers become faster and faster, it is nowadays possible to routinely apply some approaches requiring very high computing power. Two of them are Genetic Algorithms (GA) and Artificial Neural Networks (ANN). 

Some individuals can produce speech, but it is dysarthric and very hard to understand. Yet the utterance does contain information. Can this limited information be used to figure out what the individual wanted to say, and then voice it by artificial means Research labs are now employing neural network theory to determine which pauses in an utterance are due to content (i.e., between a word or sentence) and those due to unwanted halts in speech production. 

---