Research Area: Mechanical Engineering and Automation

Title: The optimization algorithm of blind walking

Abstract: The basic idea of one-dimensional blind walking algorithm is introduced. Three steps of the algorithm are introduced, the first two of which are the forward and backward method to determine the interval of the extreme point. The application of this algorithm in the multi-dimensional unconstrained optimization method is introduced, such as the broken line negative gradient direction method applied to the negative gradient direction method, and the high-order objective function requirements found in the multi-dimensional Newton method (the second-order approximate fixed-point method). The topic report consists of the following parts:

1) The basic idea of blind walking.

2) Optimization steps of one-dimensional blind walking method.

3) Negative gradient direction method of broken line.

4) Application of multidimensional Newton method.

5) Application of random direction method and polyhedron deformation method.

6) Relevant references.