The solution of simulated annealing algorithm for continuous variable optimization problem

Algorithm flow, write a little messy, see for yourself

Markov chain length is the number of iterations at a certain temperature

1, set the initial parameters: starting point, initial temperature T0, Markov chain length max_l, the maximum value of the target function max_e, the stop temperature te, the cooling function (the simplest to take linear), scale;

2, judge whether Convergence (t<te), is the word stop algorithm; otherwise t= at,l=1 on the next step;

3, randomly select a component of the current solution, according to the formula to create a disturbance, to obtain a new candidate solution, here with the upper and lower bound of the optimization problem as an example:

4, the calculation of the target function is poor, DE, according to Metropolis rules to determine whether to accept the candidate solution

Generate a random number between 0-1 R, if P>r, then accept the candidate solution as a new solution, l+=1, otherwise proceed to the next step;

5, if l>max_l, return to the third step; otherwise l+=1, return to the third step.

The solution of simulated annealing algorithm for continuous variable optimization problem