Creating and sharing knowledge for telecommunications

Online MKL for Structured Prediction

Martins, A. ; Figueiredo, M. A. T. ; Aguiar, P.

Online MKL for Structured Prediction, Proc NIPS Workshop in Optimization for Machine Learning, Whistler, Canada, Vol. ?, pp. ? - ?, December, 2010.

Digital Object Identifier:

Abstract
''Training structured predictors often requires a
considerable time selecting features or tweak-
ing the kernel. Multiple kernel learning (MKL)
sidesteps this issue by embedding the kernel
learning into the training procedure. Despite the
recent progress towards efficiency and scalabil-
ity of MKL algorithms, the structured output case
remains an open research front. We claim that the
existing wrapper-based methods are inadequate
for this task. Instead, we propose a new fam-
ily of online proximal algorithms able to tackle
many variants of MKL and group-LASSO, and
for which we show regret, convergence, and gen-
eralization bounds. Experiments on handwriting
recognition and dependency parsing illustrate the
success of the approach.''