The dream of capturing the power of the human brain using the devices of modern technology has long held strong appeal in more than academic circles. The advantages bestowed on the owner of such technology are enormous. The ubiquitousness and indispensability of the digital computer attests to its importance, yet it does not satisfy the dream. Capable of mathematical calculations far beyond the powers of mere mortals, the computer fails to match a one-yearold's pattern recognition ability or a two-year-old's motor coordination. In recent years, the flood of published research on artificial neuron-like networks-or neural nets-confirms that the dream is still alive. These relatively primitive contraptions, devoid of instruction sets, registers, and random access memory, have proved capable of "learning" to control physical systems, recognize visual and aural patterns, and perform a number of other adaptations reminiscent of nervous-system responses to its environment.
In this issue, five widely different views into current neural nets research are presented. It becomes clear from even a superficial look at the articles that artificial neural networks themselves are part of a spectrum of adaptive mechanisms vying to fulfill the dream of mimicking the biological neural network in its power to respond to environmental input in complex ways. The first paper, "Adaptive Neural Networks and their Applications" by Widrow and Lehr, provides an introduction and some of the history of the development of several widely used neural net architectures and learning (or training) algorithms. Most of the techniques discussed grew out of research conducted in Widrow's laboratory over the past two decades. From the relatively simple perceptron with Adaline elements to the more recent nonlinear Madaline Rule III, the article presents clear descriptions and motivations for both the architecture and the training rules of the various systems.
Gawronski and Rodriguez begin the second article, "A Learning Algorithm for the Classification of Dynamic Events Using a Neuron-like Dynamic Tree," with a review of some of the methodologies which have been employed