This thesis presents the use of a new sigmoid activation function in backpropagation artificial neural networks (ANNs). ANNs using conventional activation functions may generalize poorly when trained on a set which includes quirky, mislabeled, unbalanced, or otherwise complicated data. This new activation function is an attempt to improve generalization and reduce overtraining on mislabeled or irrelevant data by restricting training when inputs to the hidden neurons are sufficiently small. This activation function includes a flattened, low-training region which grows or shrinks during back-propagation to ensure a desired proportion of inputs inside the low-training region. With a desired low-training proportion of 0, this activation function reduces to a standard sigmoidal curve. A network with the new activation function implemented in the hidden layer is trained on benchmark data sets and compared with the standard activation function in an attempt to improve area under the curve for the receiver operating characteristic in biological and other classification tasks.
","abstract_html":"<p>This thesis presents the use of a new sigmoid activation function in backpropagation artificial neural networks (ANNs). ANNs using conventional activation functions may generalize poorly when trained on a set which includes quirky, mislabeled, unbalanced, or otherwise complicated data. This new activation function is an attempt to improve generalization and reduce overtraining on mislabeled or irrelevant data by restricting training when inputs to the hidden neurons are sufficiently small. This activation function includes a flattened, low-training region which grows or shrinks during back-propagation to ensure a desired proportion of inputs inside the low-training region. With a desired low-training proportion of 0, this activation function reduces to a standard sigmoidal curve. A network with the new activation function implemented in the hidden layer is trained on benchmark data sets and compared with the standard activation function in an attempt to improve area under the curve for the receiver operating characteristic in biological and other classification tasks.</p>","abstract_has_math":false,"creators":["Bonnell, Jeffrey A"],"institution":null,"degree_name":"MS (Master of Science)","degree_level":"Thesis - unrestricted","degree_discipline":"Mathematical Sciences","degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2011,"date_issued":"2011-08-17T07:00:00Z","date_published":"2011-08-17T07:00:00Z","updated_at":"2026-07-24T02:20:14Z","subjects":["neural network","machine learning","overfitting","overtraining","sigmoid","activation function","Artificial Intelligence and Robotics","Computer Sciences","Physical Sciences and Mathematics"],"languages":[],"rights":["Copyright by the authors."],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://dc.etsu.edu/etd/1342","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Bonnell, Jeffrey A"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["1990-01-01T08:00:00Z"]},{"key":"dc:date.issued","label":"Date","values":["2011-08-17T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Mathematical Sciences"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis - unrestricted"]},{"key":"thesis:degree_name","label":"Degree Name","values":["MS (Master of Science)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["neural network","machine learning","overfitting","overtraining","sigmoid","activation function","Artificial Intelligence and Robotics","Computer Sciences","Physical Sciences and Mathematics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["Copyright by the authors."]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://dc.etsu.edu/context/etd/article/2533/viewcontent/BonnellJ080311f.pdf","https://dc.etsu.edu/etd/1342"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This thesis presents the use of a new sigmoid activation function in backpropagation artificial neural networks (ANNs). ANNs using conventional activation functions may generalize poorly when trained on a set which includes quirky, mislabeled, unbalanced, or otherwise complicated data. This new activation function is an attempt to improve generalization and reduce overtraining on mislabeled or irrelevant data by restricting training when inputs to the hidden neurons are sufficiently small. This activation function includes a flattened, low-training region which grows or shrinks during back-propagation to ensure a desired proportion of inputs inside the low-training region. With a desired low-training proportion of 0, this activation function reduces to a standard sigmoidal curve. A network with the new activation function implemented in the hidden layer is trained on benchmark data sets and compared with the standard activation function in an attempt to improve area under the curve for the receiver operating characteristic in biological and other classification tasks.
"]},{"key":"dc:title","label":"Title","values":["Implementation of a New Sigmoid Function in Backpropagation Neural Networks."]}]}],"canonical_facts":{"dc:creator":["Bonnell, Jeffrey A"],"dc:date.available":["1990-01-01T08:00:00Z"],"dc:date.issued":["2011-08-17T07:00:00Z"],"dc:description.abstract":["This thesis presents the use of a new sigmoid activation function in backpropagation artificial neural networks (ANNs). ANNs using conventional activation functions may generalize poorly when trained on a set which includes quirky, mislabeled, unbalanced, or otherwise complicated data. This new activation function is an attempt to improve generalization and reduce overtraining on mislabeled or irrelevant data by restricting training when inputs to the hidden neurons are sufficiently small. This activation function includes a flattened, low-training region which grows or shrinks during back-propagation to ensure a desired proportion of inputs inside the low-training region. With a desired low-training proportion of 0, this activation function reduces to a standard sigmoidal curve. A network with the new activation function implemented in the hidden layer is trained on benchmark data sets and compared with the standard activation function in an attempt to improve area under the curve for the receiver operating characteristic in biological and other classification tasks.
"],"dc:identifier":["https://dc.etsu.edu/context/etd/article/2533/viewcontent/BonnellJ080311f.pdf","https://dc.etsu.edu/etd/1342"],"dc:rights":["Copyright by the authors."],"dc:subject":["neural network","machine learning","overfitting","overtraining","sigmoid","activation function","Artificial Intelligence and Robotics","Computer Sciences","Physical Sciences and Mathematics"],"dc:title":["Implementation of a New Sigmoid Function in Backpropagation Neural Networks."],"thesis:degree_discipline":["Mathematical Sciences"],"thesis:degree_level":["Thesis - unrestricted"],"thesis:degree_name":["MS (Master of Science)"]},"updated_at":"2026-07-24T02:20:14Z"}