Author(s)

Liu Zhengrong

Corresponding Author

Liu Zhengrong

Abstract

There are many different solutions to non smooth optimization problems. In fact, the solutions of non smooth optimization problems can be divided into two main types: sub gradient method and beam method. Any method is based on the assumption that the objective function value is unique and there is a lower gradient value at each point. The sub gradient method is a method to solve the non smooth problem based on the sub gradient. This method is simple and easy to install, but may cause convergence, optimal detection and sub gradient approximation. The connection method is the ideological basis of beam method. Based on the information obtained from the black box, a piecewise affine model of the objective function in the original problem is constructed. According to the tangent plane method, the general beam method is generated. This method is now widely used. This can keep relevant information, such as repeat points, generate sub gradients through previous repetition, and create information packages, namely bundles. Therefore, the previously established information can be used in the current iteration step to generate the next iteration point. However, with the increase of the number of repetitions, the information beam expands more and more, which leads to huge computing and storage problems. In order to solve this problem, a special integrated compression technology has appeared in beamforming.

Keywords

Composite Approach Beam Method, Linear Approximation, Dual Space, Composite Function