Support Vector Machine Implementations for Classification & Clustering
Author Information
Author(s): Stephen Winters-Hilt, Anil Yelundur, Charlie McChesney, Matthew Landry
Primary Institution: Department of Computer Science, University of New Orleans
Hypothesis
Can Support Vector Machines (SVMs) improve classification and clustering of channel current data?
Conclusion
The study demonstrates that novel, information-theoretic kernels in SVMs provide significantly better performance for classification and clustering tasks.
Supporting Evidence
- SVMs provide noise-tolerant solutions for pattern recognition.
- Novel kernels have been shown to outperform standard kernels.
- Internal multiclass SVMs improve training time without sacrificing accuracy.
- Signal clustering methods provide robust information in poorly separable data.
Takeaway
This study shows that special computer programs called Support Vector Machines can help sort and group data better than older methods.
Methodology
The study describes the use of internal and external multiclass SVMs for classification and clustering of data, employing novel kernels for improved performance.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website