CPL - Chalmers Publication Library
| Utbildning | Forskning | Styrkeområden | Om Chalmers | In English In English Ej inloggad.

Multiclass Ridge-adjusted Slack Variable Optimization Using Selected Basis for Fast Classification

Yinan Yu (Institutionen för signaler och system, Signalbehandling) ; Kostas I. Diamantaras ; Tomas McKelvey (Institutionen för signaler och system, Signalbehandling) ; S.Y. Kung
22nd European Signal Processing Conference, EUSIPCO 2014, Lisbon, Portugal, 1-5 September 2014 (2219-5491). p. 1178-1182. (2014)
[Konferensbidrag, refereegranskat]

Kernel techniques for classification is especially challenging in terms of computation and memory requirement when data fall into more than two categories. In this paper, we extend a binary classification technique called Ridge-adjusted Slack Variable Optimization (RiSVO) to its multiclass counterpart where the label information encoding scheme allows the computational complexity to remain the same to the binary case. The main features of this technique are summarized as follows: (1) Only a subset of data are pre-selected to construct the basis for kernel computation; (2) Simultaneous active training set selection for all classes helps reduce complexity meanwhile improving robustness; (3) With the proposed active set selection criteria, inclusion property is verified empirically. Inclusion property means that once a pattern is excluded, it will no longer return to the active training set and therefore can be permanently removed from the training procedure. This property greatly reduce the complexity. The proposed techniques are evaluated on standard multiclass datasets MNIST, USPS, pendigits and letter which could be easily compared with existing results.

Nyckelord: kernel; large scale data; multiclass classification; RiSVO; RKHS basis construction



Den här publikationen ingår i följande styrkeområden:

Läs mer om Chalmers styrkeområden  

Denna post skapades 2014-07-07. Senast ändrad 2015-12-07.
CPL Pubid: 200240

 

Läs direkt!

Lokal fulltext (fritt tillgänglig)