TITLE: A Sparse Signomial Model for Classification and
Regression

SPEAKER: Professor Myong K. (MK) Jeong

ABSTRACT:

Support Vector Machine
(SVM) is one of the most popular data mining tools for solving classification
and regression problems. Due to its high prediction accuracy, SVM has been
successfully used in various fields. However, SVM has the following drawbacks.  First, it is not easy to
get an explicit description of the discrimination (or regression) function in
the original input space and to make a variable selection decision in the input
space. Second, depending on the magnitude and numeric range of the given data
points, the resulting kernel matrices may be ill-conditioned, so learning
algorithms may be suffered from numerical instability even though data scaling
generally helps to handle this kind of issues but may not be always effective.
Third, the selection of an appropriate kernel type and its parameters can be
complex while the performance of the resulting functions is heavily influenced.

To overcome these
drawbacks, this talk presents the sparse
signomial classification and regression (SSCR) model. SSCR seek a sparse
signomial function by solving a linear program to minimize the weighted sum of
the ℓ1-norm of the coefficient vector of the function and the ℓ1-norm
of violation (or loss) caused by the function. SSCR can explorevery high
demensional feature spaces with less sensitivity to numerical values or numeric
ranges of the given data. Moreover, this method give an explicit description of
the resulting function in the original input space, which can be used for
prediction purposes as well as interpretation purposes. We present a practical
implementation of SSCR based on the column generation and explore some
theoretical properties of the proposed formulation. Computational study shows
that SSCR is competitive or even better performance compared to other widely
used learning methods for classification and regression.