We give a novel proof of the $O(1/k)$ and $O(1/k^2)$ convergence rates of theproximal gradient and accelerated proximal gradient methods for compositeconvex minimization. The crux of the new proof is an upper bound constructedvia the convex conjugate of the objective function.
展开▼
机译：我们给出了新的$ O（1 / k）$和$ O（1 / K ^ 2）$收敛率的新颖证明，并加速了CompositeConvex最小化的近端梯度方法。新证据的关键是上限构建维亚目标函数的凸缀合物。
展开▼
