Adaptive Processing at Cheshire Engineering CorporationCheshire has developed capabilities at the forefront of new adaptive signal processing technologies, particularly in the area of neural networks and genetic algorithms.
Neural networks are self-organized, self-learning data processing networks which can be "trained" to perform an extremely broad class of functions. Neural networks are even more powerful in that they can adapt to changing patterns and conditions after initial "training". Neural networks were first developed as adaptive signal processing algorithms in the 1960's and 1970's. The theoretical framework was reorganized in the 1980's and the field was given the present name of neural networks. Neural networks are becoming more ubiquitous, and have seen applications in: pattern and object recognition, trend prediction, knowledge-base systems, and other like applications.
Genetic algorithms are another class of new techniques used to create self-organizing, adaptive systems. Genetic algorithms work by defining a key set of controlling parameters for a base algorithm. Then by "evolving" those controlling parameters, using algorithmic analogs to biological evolution, improved algorithmic performance is achieved.
In addition to developing and applying this expertise to client applications, Cheshire has also written commercial programs implementing these techniques as a tool package for other users. Further information on neural networks and Cheshire's Neuralyst Products is available.