This paper presents a coordinate gradient descent approach for minimizing the sum of asmooth function and a nonseparable convex function. We ¯nd a search direction by solvinga subproblem obtained by a second-order approximation of the smooth function and addinga separable convex function. Under a local Lipschitzian error bound assumption, we showthat the algorithm possesses global and local linear convergence properties. We also givesome numerical tests (including image recovery examples) to illustrate the e±ciency of theproposed method.
展开▼