Support Vector Machine Training by Mechanical Equilibrium

Resumo: The Support Vector Machine (SVM) is currently the most successful classification architecture, used in the most diverse fields of application. The orthodox framework to train a Support Vector Machine is constraint optimization based on Convex Quadratic Programming (QP).
This approach quickly exhausts the memory resources of a digital computer when the number of training samples is growing and is usually circumvented by introducing sub-optimal heuristics. It is known that a trained SVM is in a mechanical equilibrium of forces and torques when considering that each pattern exerts a perpendicular force on the separating hyperplane. By now, no practical algorithm has been
proposed that leads exactly to this state of equilibrium by using only unconstrained gradient descent optimization. The benefits of this new way to train a SVM are a considerable reduction of RAM memory and CPU time, thus avoiding heuristics.

Data de início: 2011-03-01
Prazo (meses): 24

Participantes:

Papelordem decrescente Nome
Coordenador Thomas Walter Rauber
Acesso à informação
Transparência Pública

© 2013 Universidade Federal do Espírito Santo. Todos os direitos reservados.
Av. Fernando Ferrari, 514 - Goiabeiras, Vitória - ES | CEP 29075-910