erpretation, but use binary processing to arrive at decisions. The problem is determining at what point a neural network is appropriate, and at what point an expert system should be used.
The basic difference between expert systems and neural networks is that expert systems use analytical abstraction, while neural networks rely on binary processing. Developers of neural networks learned that biological neurons, similar to computer bits, are either on or off. Neural networks try to reproduce the biological network electronically to approximate the abilities of the brain and nervous system to recognize patterns, satisfy constraints and process signals. The biological brain consist of 100 billion neurons, each connected to 10,000 others by synapses; neural networks are subject to size constraints and can thus be constructed for limited applications, at this point.
Despite the constraints, neural networks are constructed in such a way that they correspond approximately to the way that people learn. A network is built of "neurons" that make decisions, classifications and forecasts. the system is trained by the repeated display of previously solved examples that contain explicit data sets if facts or parameters with corresponding answers or results. The input of a large amount of data (as many as 50,000 examples) is required to establish a reliable neural network (May, 1991, p. 118).
The value of neural networks is in their ability to classify information or predict results of new input based on experience. This is the machine's ability to "think." New situations are analyzed in light of previous situations that the machine has encountered in the past. This is done by the network placing a weighted value on inputs and passing that value to the next neuron, which is activated (or not) based on the value passed to it. Neural networks learn by adjusting the interco
...