"Optimization Algorithms for Deep Learning" (18) Brainstorming Algorithms

1. Introduction to brainstorming algorithm

(The following descriptions are not academic terms, just for your happy reading.)
  Maybe you are not very familiar with the word "brainstorming". After all, it is a foreign word, and its general meaning is to discuss in groups and speak freely .
  Brainstorming algorithm (Brain Storm Optimization) is an optimization algorithm proposed according to the process of people "brainstorming" to discuss solutions to difficult problems. The algorithm was proposed in 2011. I am very happy that this is another optimization algorithm proposed by a Chinese. However, the description and implementation of this algorithm are relatively complicated, and there are not many related papers. You can try to "develop" it.
  In the brainstorming algorithm, each person's position (idea, described as a better understanding of the position) represents a feasible solution in the solution space. The group will be divided into several groups for their brainstorming, and the person with the best position in each group is called the leader of the group, also called the cluster center. Each individual has two ways to reach a new position: thinking alone and thinking together , and the leader of each group has a new way: thinking randomly . Everyone will choose their way of moving according to certain rules, and choose whether to update to this new idea. The specific implementation is more complicated and will be described in detail in the next section.

2. Algorithm process

  The number of individuals in the crowd is N, its idea (position, in fact, it is more convenient to use position to describe) is, and its fitness value is .
The brainstorming algorithm is roughly divided into two steps

Guess you like

Origin blog.csdn.net/quanzhan_King/article/details/131863363