Abstract
This paper presents a novel method for dimensionality reduction and for multi-class classification tasks. This method iteratively selects a series of simple but effective 1D subspaces, and then combines the corresponding 1D projections by Adaboost. M2. Its major advantages are: 1) it does not impose speci.c assumptions on data distribution; 2) it minimizes Bayes error estimation in low-dimensional space; 3) it simpli.es existing subspace-based methods to eigenvalue decomposition problem; and 4) each of the 1D subspaces (with associated nearest neighbor classi.er) has different emphasis - measured by weighted training error. Experiments on both synthetic and real-world data demonstrate the effectiveness of the proposed method.
Original language | English |
---|---|
Article number | 4811624 |
Pages (from-to) | 2231-2235 |
Number of pages | 5 |
Journal | Conference Proceedings - IEEE International Conference on Systems, Man and Cybernetics |
DOIs | |
State | Published - 2008 |
Externally published | Yes |
Event | 2008 IEEE International Conference on Systems, Man and Cybernetics, SMC 2008 - Singapore, Singapore Duration: 12 Oct 2008 → 15 Oct 2008 |
Keywords
- Adaboost.M2
- Feature extraction
- Multi-class classification