Physical Sciences and Mathematics Commons^™

Neural Network Classification Of Environmental Samples, Jeffrey L. Blackmon

Theses and Dissertations

This research develops a general methodology for designing neural network classifiers for real-world environmental problems. This methodology is demonstrated through the design of a multi-layer perceptron to classify stainless steel and actinide samples. Neural networks, sometimes called artificial neural networks, have been shown capable of classifying complex patterns. Artificial neural networks are physiologically motivated computer algorithms which attempt to mimic the function of the large interconnected network of neurons in the human brain, which has extraordinary pattern recognition capabilities. These artificial neural networks learn to map a set of input features, elemental composition, onto a set of outputs such as …