|
What Is a Neural Network ? Neural networks have the remarkable ability to solve problems related to detecting trends and patterns that humans or other computer techniques are unable to solve. An Artificial Neural Network, (or an ANN), is an information-processing paradigm that models the way in which biological nervous systems, like the brain, process information. Contrary to computers, the brain consists of billions of processors, which process a large number of tasks concurrently. Neurons work collaboratively to solve the defined problem. Simulating these processes of the brain, neural networks are adept in areas that resemble human reasoning, making them well suited to solve problems that involve pattern recognition and forecasting. Structure An ANN consists of two primary parts: neurons, represented by neural units, and synapses, connections between the neurons, which send signals from neuron to neuron. Those synapses can be: Excitated (positive weight) or Inhibited (negative weight). Architecture Most known neural networks have at least two layers: 1) input layers relate to what the agent receives from the environment, and 2) output layers, which correspond to the agent's potential actions. Others (like Back Propagation) have one or more intermediate layers between these two layers. These layers are highly interconnected, as the units on one layer are connected to those in the next layer. Shaping Factors Just like the factors that shape a human, the factors that shape a neural network are its environment and its genetic makeup. Both its initial state and its training play a role in the ANN's development. It is through the critical training process that ANN's are taught how to arrive at the correct answer. This training process works much in the same way that a child comes to learn, for example, not to eat a particular food that repeatedly causes an upset stomach. Through different samples and experiences, the child eventually stops eating this food. So clearly, a well-trained neural network will be more successful than a poorly trained neural network (the training referring to its environment and the experiences and samples that help shape it). The more samples (experiences) to which a neural network has a direct correlation, the greater the likelihood of its success.
Deciding on a Neural Network Choosing which neural network to use for a particular application is much like deciding how to furnish an apartment. With many approaches that can be pursued, determining the right one can be difficult: one might be to choose a furnishing that complements the time-period of the apartment, decorating it to accent the historical qualities. A different approach might be economical, selecting minimal furniture to complement a tighter budget, thus acknowledging a financial condition. Yet another approach could be to settle on a standard solution, one that does not necessarily make a statement or respond to any aspect of the apartment -- a tried and true, run of the mill style. Companies often take this last approach, using Back Propagation to solve many problems, eliminating the option of a more appropriate neural network. Generally, several neural networks can be applied to a particular application, but there is usually a best choice. Limitations of Neural Network Technically Neural networks are algorithmic systems which interpret historical data to identify trends and patterns against which to compare subject cases. When a case is presented, its factors are fed into the input layer of the neural net. The factors’ weights are adjusted as the numbers feed through the neural net and reach the output nodes as a single source accompanied by a certainty factor. Typical neural networks are limited by their training: i.e. they can reliably recognize only the kind of information on which they were trained. New information will be patiently processed, but the result provided can be nonsense.Brighterion's Neural Network Suite iPrevent’s proprietary neural network can translate any database to neurons without user intervention and has significantly accelerated the speed of convergence as compared to typical neural network algorithms such as backpropagation. Brighterion’s neural net is incremental and adaptive, allowing the size of the output classes to change dynamically. Additionally, in its expert mode, iPrevent provides a library of twelve different neural network models for use in customization. |