Abstract
Nonquadratic regularizers, in particular the l1 norm regularizer can yield sparse solutions that generalize well. In this work we propose the generalized subspace information criterion (GSIC) that allows to predict the generalization error for this useful family of regularizers. We show that under some technical assumptions GSIC is an asymptotically unbiased estimator of the generalization error. GSIC is demonstrated to have a good performance in experiments with the l1 norm regularizer as we compare with the network information criterion (NIC) and cross- validation in relatively large sample cases. However in the small sample case, GSIC tends to fail to capture the optimal model due to its large variance. Therefore, also a biased version of GSIC is introduced, which achieves reliable model selection in the relevant and challenging scenario of high-dimensional data and few samples.
Original language | English |
---|---|
Pages (from-to) | 70-80 |
Number of pages | 11 |
Journal | IEEE Transactions on Neural Networks |
Volume | 13 |
Issue number | 1 |
DOIs | |
Publication status | Published - 2002 Jan |
Keywords
- Kernel methods
- Model selection
- Regularization
- Sparse regressors
- Subspace information criterion (SIC)
ASJC Scopus subject areas
- Software
- Computer Science Applications
- Computer Networks and Communications
- Artificial Intelligence