NeuralNex.com
Introduction to Neural Networks

NeuralNex.com is currently an informational site on next generation neural network computing. Neural networks learn by exposure to examples, not by well-defined programming. In some respects, this is like the apprentice function in a large number of human professions. People aspiring to a given finance first observe, and then participate in, several cases done by a master. Eventually, their period of apprenticeship is over and they become a master. The apprentice period for neural networks is labeled "supervised practice." They "observe" cases and adjust their internal configuration so as to best proliferate the outcomes desired by the "master" - the human supervisor. The number of cases that is necessitated before they master a problem depends on the complexity of the problem and the variability in case characteristics. The more complicated the case and the more pushing parts that must be "sorted through" to observe the critical underlying connections, the more cases are needed. Artificial Neural Networks are trained to be experts in solving particular types of problems by training on cases in which the right outcome is known. The education cases are used to adjust the network's weights and configuration so as to maximize its exactness in selecting the right outcomes. There must be the right amount of cases so that the network can construct an accurate prototype linking inputs to the desired outputs without over-fitting the files. With insufficient cases, the simulation may not converge. Once trained, the network is then practical to cases where the right outcomes are not known. Neural networks are particularly good at recognizing patterns that conventional computers will likely not. Artificial neural networks, patterned after their living counterparts, learn from interaction with their environment. The connections among their processing parts are constantly shifting. Connections that contribute to successful results in environmental interactions are reinforced. Connections that contribute to failures in ecological interactions are reduced and might eventually disconnect. This is analogous to the dynamic adjustments to synaptic connections among neurons in organic neural networks. Adaptive neural growth is the vital to education in both organic and inorganic neural networks. Neural networks begin as a flexible, highly-randomized array of densely inter-connected signal processing neural units. Interaction between the neural network and its environment varies the relative strength and boom of inter-connections and nodes. Connections that are confirmed by subsequent natural interactions are reinforced and grow. Connections that are not confirmed are weakened and atrophy. This creates an adaptive loop between neural growth and environmental realities. Click on images to learn more about biotech and neurotech T-shirts: Questions and suggestions concerning the site NeuralNex.com may be sent to: NeuralNex.com © 2006 by NeuralNex.com