SVDD (Support Vector Data Description) Algorithm in LIBSVM
- Login to Download
- 1 Credits
Resource Overview
SVDD (Support Vector Data Description) algorithm in LIBSVM can be directly implemented and allows control over the number of remaining support vectors by adjusting gamma and C parameters through configuration file settings or function arguments.
Detailed Documentation
The SVDD (Support Vector Data Description) algorithm in LIBSVM is a widely-used anomaly detection method that constructs an optimal hypersphere to encapsulate normal data patterns while identifying outliers. This implementation supports direct usage with configurable parameters: gamma controls the kernel function's influence range (e.g., RBF kernel width), while parameter C regulates the trade-off between model complexity and training error tolerance. Key functions include svmtrain() for model building with 'svdd' type specification, and svmpredict() for anomaly scoring. The algorithm's flexibility enables tuning support vector quantities via parameter adjustment, making it adaptable to diverse anomaly detection scenarios through systematic hyperparameter optimization.
- Login to Download
- 1 Credits