Sadanori Konishi
Efficient algorithm to select tuning parameters in sparse regression modeling with regularization.
Fri, 09/16/2011 - 15:01 — SomNambulAuthors: Sadanori Konishi, Kei Hirose, Shohei Tateishi
Subjects: Methodology
AbstractIn sparse regression modeling with regularization such as the lasso, elastic
net and bridge regression, it is important to select appropriate values of
tuning parameters including regularization parameters. The choice of tuning
parameters can be viewed as a model selection and evaluation problem. The
degrees of freedom, which leads to Mallows' $C_p$ criterion, plays a key role
in the theory of model selection. In the present paper, we propose an efficient
algorithm which computes the degrees of freedom sequentially by extending the
generalized path seeking (GPS) algorithm.Semi-supervised logistic discrimination for functional data.
Wed, 02/23/2011 - 16:01 — SomNambulAuthors: Shuichi Kawano, Sadanori Konishi
Subjects: Methodology
AbstractMulti-class classification methods based on both labeled and unlabeled
functional data sets are discussed. We present semi-supervised logistic models
for classification in the context of functional data analysis. Unknown
parameters in our proposed models are estimated by regularization with the help
of EM algorithm. Crucial points in modeling procedure are the choices of
regularization parameter involved in the semi-supervised functional logistic
models. In order to select the adjusted parameter, we introduce model selection
criteria from information-theoretic and Bayesian viewpoints.
User login
