Despite the recent progress towards efficient multiple kernel learning (MKL),the structured output case remains an open research front. Current approachesinvolve repeatedly solving a batch learning problem, which makes theminadequate for large scale scenarios. We propose a new family of onlineproximal algorithms for MKL (as well as for group-lasso and variants thereof),which overcomes that drawback. We show regret, convergence, and generalizationbounds for the proposed method. Experiments on handwriting recognition anddependency parsing testify for the successfulness of the approach.
展开▼