Abstract
Neural Networks are an effective means of classifying data; however they are usually purpose built applications that are created for classifying a single data set. Programming a neural network can be a time consuming and sometimes error prone process. To alleviate both of these problems a self-configuring multilayer perceptron model was used to create and train neural networks. This application can take any training data set that is linearly or nonlinearly separable as input, then create the needed neural network structure and train itself, thus saving programmers' time and effort. The software has been tested with several data sets including sample data sets, the Iris data set, and the MARSI data set. The results indicate that once the network is created and trained, it can be used to effectively classify data from many data sets.
Date of publication
Spring 3-13-2013
Document Type
Thesis
Language
english
Persistent identifier
http://hdl.handle.net/10950/105
Recommended Citation
Anderson, Justin M., "Self-Configuring Neural Networks" (2013). Computer Science Theses. Paper 2.
http://hdl.handle.net/10950/105