Traditional optimization algorithm VS intelligent optimization algorithm

Advantages of traditional optimization algorithms

1: Use the characteristics of understanding space, such as differentiability.
2: The theory is relatively complete, and the amount of calculation is small.
3: Fast convergence speed.
4: Has a definite termination criterion.

Disadvantages of traditional optimization algorithms

1: Only the local optimal solution of the optimization problem can be found.
2: The result of the solution strongly depends on the initial value.

Advantages of genetic algorithm:

1: Able to find the global optimal solution of the optimization problem.
2: The optimization result has nothing to do with the initial conditions.
3: The algorithm is independent of the solution domain.
4: It has strong robustness (anti-interference).
5: Suitable for solving complex optimization problems.
6: It is widely used.

Disadvantages of genetic algorithm

1: The convergence speed is slow.
2: Poor local search capability.
3: Many control variables.
4: There is no definite termination criterion.

Comparison of features

1. Traditional optimization algorithms typically target structural problems, there are more describe specific conditions and problems , such as linear programming, quadratic programming, integer programming, mixed planning with constraints and the like without constraints, i.e. have a clear structural information ; And the intelligent optimization algorithm generally aims at a more general problem description, and generally lacks structural information.

2. Many traditional optimization algorithms belong to the category of convex optimization, and have the only clear global optimum ; while the vast majority of intelligent optimization algorithms are aimed at multi-extreme problems . How to prevent falling into local optimum and find the global optimum as much as possible is to adopt intelligence The root cause of the optimization algorithm:

3. For single-extreme problems, traditional algorithms are good enough most of the time, but smart algorithms do not have any advantages; for multi-extreme problems, intelligent optimization algorithms can have a difference between jumping out of the local optimum and converging to a point through its effective design. Better balance, so as to find the global optimum, but sometimes the local optimum is also acceptable, so traditional algorithms also have a lot of application space and improvement possibilities for special structures.

4. Traditional optimization algorithms are generally deterministic algorithms with fixed structures and parameters. The computational complexity and convergence can be theoretically analyzed; intelligent optimization algorithms are mostly heuristic algorithms, which can be qualitatively analyzed but difficult to quantitatively prove, and most algorithms are based on random Characteristics, its convergence is generally in the sense of probability, the actual performance is uncontrollable, the convergence speed is often slow, and the calculation complexity is high.
   btw, your teacher’s original saying "there is no way" is basically correct. If there is a definite description of the problem structure and the corresponding traditional algorithm (theoretical proof can guarantee its reliability), who will choose the calculation and the complicated result What about uncontrollable intelligent algorithms? Only when it is difficult for traditional algorithms to get an acceptable optimal solution (in many cases in practical problems, global optimality is not required, and intelligent algorithms do not guarantee 100% to find global optimality), intelligent algorithms are indeed an option. There is no way. Why is the research on intelligent algorithms more popular, because the former is hard bones and requires strong mathematical skills to make small progress, while intelligent algorithm undergraduates can write articles with their heads—although high-quality articles It is still difficult.