Bundle Methods for Machine Learning (T17)
Alex Smola, S V N Vishwanathan, Quoc Viet Le

· Problem: Regularized Risk Minimization
· Our Idea: Use modified Bundle Method Solver
 convergence rates for non-smooth obj.
 convergence rates for smooth obj.  Our algorithm applies to a wide variety of loss fn.  SVMPerf a special case of our method

Objective Function vs. No. of Iterations