当前位置:网站首页>Machine learning notes - gray wolf optimization

Machine learning notes - gray wolf optimization

2022-07-05 14:46:00 Sit and watch the clouds rise

1、 summary

         Among various optimization techniques , Gray wolf optimization is a meta heuristic optimization technology , Its inspiration comes from the hierarchical relationship between wolf families and the special hunting techniques used by gray wolves . Therefore, gray wolf optimization technology simulates the overall characteristics of gray wolf population , Try to find the optimal solution .

         Before understanding the optimization technology of gray wolf , Let's try to understand why this algorithm is inspired by the social hierarchy of the gray wolf family .

         The above figure shows the social hierarchy of gray wolves , The characteristics of each wolf category are different in the Group . The whole gray wolf optimization family is officially called a pack. Now let's try to understand the responsibilities of each kind of wolf in the Group .

         Alpha wolf : Alpha wolves dominate the gray wolf pack , Have the right to command the whole gray wolf group .

        Beta Wolf: Beta Wolves regularly give Alpha Wolf Report , And to help Alpha The wolf makes the best decision .

        Delta Wolf: Delta The wolf belongs to beta The Wolf , by alpha and beta Wolf provides continuous updates , yes omega The wolf's superior .

         Omega wolf : Omega wolves are responsible for hunting wolves in gray wolves , And take care of young wolves .

         Gray wolves follow a special hunting technique , The whole gray wolf pack hunts its prey in groups . The selected prey is separated from the pack by Omega wolves , The selected prey is chased and attacked by coyotes and beta wolves . therefore , The unique hunting technology adopted by gray wolves has led to the development of an optimization technology called gray wolf optimization , Among them, the use of various built-in functions produces the closest optimal solution .

2、 purpose

         Gray wolf optimization technology is used for various time-consuming problems , for example NP-hard Problems and travel sales problems . Gray wolf optimization technology usually reduces the operation time of high-dimensional data , Because the algorithm decomposes the whole complex problem into subsets . Provide a subset of operations to each agent , Similar to the overall hierarchy of gray wolves , And produce the best optimal solution .

         therefore , In an algorithm similar to gray wolf hierarchy , Decompose the complex problem into agents , Each agent undertakes its own tasks and reduces the overall time consumption . All agents in the algorithm follow certain guidelines and strategies , And find the best solution of the problem . 

3、 working principle

         Explain the gray wolf pack in the form of optimal solution parameters .

        Alpha wolf It can be called the most suitable solution of all possible solutions to this problem . It is the optimal solution produced by the optimization algorithm .

        Beta wolf It can be called the sub optimal solution of all possible solutions to the problem . If the optimal solution is not suitable for some solutions , Then the solution will be adopted .

        Delta wolf It can be called the third best solution of all possible solutions to this problem . But for all possible solutions , The third best solution is to use the most suitable and most suitable solution to evaluate .

        Omega wolf It can be called the optimal solution generated for all possible solutions , And only the third optimal solution is used to evaluate the optimal solution , And will not be compared with the best appropriate solution .

         In the context of the optimal solution generated by gray wolf optimization technology , Let's try to understand how the solution generated by the algorithm is listed as the most suitable solution among all possible solutions .

         First , Verify a random number of possible solutions to the problem . All possible solutions are usually expressed as “A” The standard vector proportion of . therefore , If A>1, Then the possible solution will deviate from the optimal solution of the problem , If A<1, Then all possible solutions will converge to the optimal solution , To find the most appropriate solution to the problem . Once the most suitable solution is determined , The algorithm will stop iterating , And appropriately rank the best possible solutions to the problem , And get the verified solution from the ranking . in the majority of cases , Use the most suitable solution , In rare cases , Choose the next best solution to some problems , Not the most suitable solution .

4、 Application example

         Gray wolf optimization finds its main application in tasks that must repeat the solution to achieve the required tasks to produce the best solution . therefore , Gray wolf optimization technology has been applied in various problems , for example NP-hards problem 、 Traveling salesman problem and many others AI problem . Let's try to understand how grey wolf optimization technology can help solve the problem of travel agents .

         Let's first try to understand the problem of travel agents . The goal of this problem is that the salesperson is only allowed to visit the city once , Find the shortest path between cities .

         Explain the traveling salesman problem through gray wolf optimization , The whole gray wolf population can be called the different paths that salesmen must take to cover the city , Therefore, it becomes a candidate solution to this problem . The best solution can be taught as prey , The best solution here is the best starting city . 

         So for the starting city , The optimization algorithm will have to produce the nearest city ( Alpha wolf ), This can be called the best solution to the problem , Therefore, all cities that are the second closest to the starting city can be called the second best solution (beta wolf). Similarly , The city that is the third closest to the starting city can be called the third best solution (delta wolf), And all other cities closest to the smart city can be called the best solution among all candidate solutions (omega wolf ). 

         The most suitable solution is TSP The best solution for the solution , Known as the best fit solution , And the best fitting solution is verified in all the best solutions generated by gray wolf optimization .

         So this is how the gray wolf optimization algorithm will find the best solution for salesmen by providing the shortest distance from the starting city to the next city and ensuring that salesmen only visit each city once .

5、 Advantages and disadvantages

         Advantages of grey wolf optimization

         Because gray wolf optimization technology attempts to replicate the hunting characteristics of gray wolves , The optimization algorithm decomposes the complex problem into different subsets , And try to produce the best possible optimal solution . Compared with other optimization algorithms , The iteration process of gray wolf optimization algorithm is faster , Because they compare different solutions for the best solution and sort them accordingly . This sort of grey wolf optimization technology makes the convergence speed of the model faster .

         Disadvantages of gray wolf optimization

         Gray wolf optimization only tries to find the best solution when the best possible solution falls within the scope of the best solution . This makes gray wolf optimization technology produce low accuracy , Sometimes it converges to a bad solution . In some cases , The best possible solution may not be considered by the candidate solution group . Besides , Gray wolf optimization technology belongs to heuristic optimization technology , The resulting optimal solution is only close to the original optimal solution , It is not the best solution of the problem .

原网站

版权声明
本文为[Sit and watch the clouds rise]所创,转载请带上原文链接,感谢
https://yzsam.com/2022/186/202207051443165471.html