Structured statistical learning with Support Vector Machine
for feature selection and prediction
Yoonkyung Lee
Ohio State
The Support Vector Machine (SVM) has been a popular choice of
classification method for many applications in machine learning.
Despite its competitive classification accuracy, the implicit nature
of the solution renders the SVM less attractive in providing insights
into the relationship between covariates and classes.
To enhance interpretability of the SVM, we integrate feature selection
into the general setup of nonlinear SVM for multiclass case. Simultaneous
feature selection and prediction is accomplished by functional ANOVA
decomposition and the L_1 penalty imposed on the scale parameters of
functional subspaces. In addition, characterization of the solution path of
the multicategory SVM is discussed as a computational shortcut to attain
the
entire spectrum of solutions from the most regularized to the completely
overfitted ones.