Multiclass Latent Locally Linear Support Vector Machines
| Type of publication: | Conference paper |
| Citation: | Fornoni_ACML2013_2013 |
| Publication status: | Published |
| Booktitle: | JMLR W&CP, Volume 29: Asian Conference on Machine Learning |
| Year: | 2013 |
| Pages: | 229-244 |
| Location: | Canberra, Australia |
| ISSN: | 1938-7228 |
| URL: | http://jmlr.org/proceedings/pa... |
| Abstract: | Kernelized Support Vector Machines (SVM) have gained the status of off-the-shelf classifiers, able to deliver state of the art performance on almost any problem. Still, their practical use is constrained by their computational and memory complexity, which grows super-linearly with the number of training samples. In order to retain the low training and testing complexity of linear classifiers and the exibility of non linear ones, a growing, promising alternative is represented by methods that learn non-linear classifiers through local combinations of linear ones. In this paper we propose a new multi class local classifier, based on a latent SVM formulation. The proposed classifier makes use of a set of linear models that are linearly combined using sample and class specific weights. Thanks to the latent formulation, the combination coe#cients are modeled as latent variables. We allow soft combinations and we provide a closed-form solution for their estimation, resulting in an effcient prediction rule. This novel formulation allows to learn in a principled way the sample specific weights and the linear classifiers, in a unique optimization problem, using a CCCP optimization procedure. Extensive experiments on ten standard UCI machine learning datasets, one large binary dataset, three character and digit recognition databases, and a visual place categorization dataset show the power of the proposed approach. |
| Keywords: | Latent SVM, Locally Linear Support Vector Machines, multiclass classification |
| Projects |
Idiap |
| Authors | |
| Editors | |
| Added by: | [UNK] |
| Total mark: | 0 |
|
Attachments
|
|
|
|
|
Notes
|
|
|
|
|
