Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
Queen's University Belfast
A sequential algorithm for sparse support vector classifiers
This paper proposes a new algorithm for training support vector machines. When compared with the best algorithms currently available using three public benchmark problems and a human activity recognition application our technique has competitive accuracy whilst maintaining faster and more stable running time. These algorithms formed part of a toolkit developed for the HaptiMap project (FP7-ICT-224675, 7.7MEuro) that aims to deeply embed accessibility into digital maps. With 13 consortium partners, the success of this project is reflected with 129 scientific papers published and 28,000 downloads of the demonstrator applications, one of which depends on the algorithms developed in this paper.