Incremental Evolution of Trainable Neural Networks that are Backwards Compatible

Simple item page
dc.contributor.authorChristenson, Chris
dc.contributor.authorKaikhah, Khosrow
dc.date.accessioned2012-02-24T10:17:47Z
dc.date.available2012-02-24T10:17:47Z
dc.date.issued2006-02
dc.description.abstractSupervised learning has long been used to modify the artificial neural network in order to perform classification tasks. However, the standard fully-connected layered design is often inadequate when performing such tasks. We demonstrate that evolution can be used to design an artificial neural network that learns faster and more accurately. By evolving artificial neural networks within a dynamic environment, the artificial neural network is forced to use learning. This strategy combined with incremental evolution produces an artificial neural network that outperforms the standard fully-connected layered design. The resulting artificial neural network can learn to perform an entire domain of tasks, including those of reduced complexity. Evolution alone can be used to create a network that performs a single task. However, real world environments are dynamic and thus require the ability to adapt to changes.
dc.description.departmentComputer Science
dc.formatText
dc.format.extent6 pages
dc.format.medium1 file (.pdf)
dc.identifier.citationChristenson, C. & Kaikhah, K. (2006). Incremental Evolution of Trainable Neural Networks that are Backwards Compatible. Proceedings of the 5th IASTED International Conference on Artificial Intelligence and Applications (AIA), pp. 222-227.
dc.identifier.urihttps://hdl.handle.net/10877/3809
dc.language.isoen
dc.publisherInternational Association of Science and Technology for Development
dc.sourceFifth IASTED International Conference on Artificial Intelligence and Applications (AIA), 2006, Innsbruck, Austria.
dc.subjectincremental evolution
dc.subjectneural networks
dc.subjecttraining
dc.subjectbackwards compatible
dc.subjectComputer Science
dc.titleIncremental Evolution of Trainable Neural Networks that are Backwards Compatible
dc.typeArticle

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
fulltext.pdf
Size:
166.25 KB
Format:
Adobe Portable Document Format
Download

Collections

College of Science and Engineering