A Parallel Mixture of SVMs for Very Large Scale Problems
Abstract
Support Vector Machines (SVMs) are currently the state-of-the-art models for many classification problems but they suffer from the complexity of their train(cid:173) ing algorithm which is at least quadratic with respect to the number of examples. Hence, it is hopeless to try to solve real-life problems having more than a few hundreds of thousands examples with SVMs. The present paper proposes a new mixture of SVMs that can be easily implemented in parallel and where each SVM is trained on a small subset of the whole dataset. Experiments on a large benchmark dataset (Forest) as well as a difficult speech database, yielded significant time improvement (time complexity appears empirically to locally grow linearly with the number of examples) . In addition, and that is a surprise, a significant improvement in generalization was observed on Forest.
Cite
Text
Collobert et al. "A Parallel Mixture of SVMs for Very Large Scale Problems." Neural Information Processing Systems, 2001.Markdown
[Collobert et al. "A Parallel Mixture of SVMs for Very Large Scale Problems." Neural Information Processing Systems, 2001.](https://mlanthology.org/neurips/2001/collobert2001neurips-parallel/)BibTeX
@inproceedings{collobert2001neurips-parallel,
title = {{A Parallel Mixture of SVMs for Very Large Scale Problems}},
author = {Collobert, Ronan and Bengio, Samy and Bengio, Yoshua},
booktitle = {Neural Information Processing Systems},
year = {2001},
pages = {633-640},
url = {https://mlanthology.org/neurips/2001/collobert2001neurips-parallel/}
}