ICML Discuss
Projection-free Online Learning
by Elad Hazan, Satyen Kale at ICML 2012
We present efficient online learning algorithms that eschew projections in favor of linear optimizations using the Frank-Wolfe technique. We obtain regret bounds that vary from the optimal $\tilde{O}(\sqrt{T})$ for stochastic online smooth convex optimization to $\tilde{O}(T^{3/4})$ for general online convex optimization. Besides the computational advantage, other desirable features of our algorithms are that they are parameter-free in the stochastic case and produce sparse decisions.

Related Material

Download PDF Watch Video

Discussion

Email notifications of comments are sent to authors.
Please use the feedback page to report broken links and other problems.
blog comments powered by Disqus