This thesis is devoted to designing efficient optimization algorithms for machine learning (ML) problems where the underlying objective function to be optimized is convex but not necessarily differentiable. Such non-smooth ...