Sparse Probability Regression by Label Partitioning

Abstract

A large-margin learning machine for sparse probability regression is presented. Unlike support vector machines and other forms of kernel machines, nonlinear features are obtained by transforming labels into higher-dimensional label space rather than transforming data vectors into feature space. Linear multi-class logistic regression with partitioned classes of labels yields a nonlinear classifier in the original labels. With a linear kernel in data space, storage and run-time requirements are reduced from the number of support vectors to the number of partitioned labels. Using the partitioning property of KL-divergence in label space, an iterative alignment procedure produces sparse training coefficients. Experiments show that label partitioning is effective in modeling non-linear decision boundaries with same, and in some cases superior, generalization performance to Support Vector Machines with significantly reduced memory and run-time requirements.

Cite

Text

Chakrabartty et al. "Sparse Probability Regression by Label Partitioning." Annual Conference on Computational Learning Theory, 2003. doi:10.1007/978-3-540-45167-9_18

Markdown

[Chakrabartty et al. "Sparse Probability Regression by Label Partitioning." Annual Conference on Computational Learning Theory, 2003.](https://mlanthology.org/colt/2003/chakrabartty2003colt-sparse/) doi:10.1007/978-3-540-45167-9_18

BibTeX

@inproceedings{chakrabartty2003colt-sparse,
  title     = {{Sparse Probability Regression by Label Partitioning}},
  author    = {Chakrabartty, Shantanu and Cauwenberghs, Gert and Jayadeva, },
  booktitle = {Annual Conference on Computational Learning Theory},
  year      = {2003},
  pages     = {231-242},
  doi       = {10.1007/978-3-540-45167-9_18},
  url       = {https://mlanthology.org/colt/2003/chakrabartty2003colt-sparse/}
}