A modified inertial three-term conjugate gradient method for nonsmooth convex optimization and its application
摘要
An inertial three-term conjugate gradient algorithm is proposed for addressing nonsmooth problem by combining inertial extrapolation step and nonmonotonic line search technique. The method presented generates ample descent directions in every iteration. Given suitable conditions, global convergence of this method is assured. Finally, it is applied to image restoration and machine learning model to illustrate effectiveness in practical scenarios.