The military-industrial scientific research system of the academic master

Fortunately, Xu Yang still remembered his own job. He was shocked and did not continue to ask questions. Instead, he quickly devoted himself to the development of parallel computing programs after seeing Chang Haonan's optimization ideas.

Since the parametric modeling, sample point selection and establishment of the kriging surrogate model had been completed before, there was not much work left for Chang Haonan and Xu Yang to complete on site.

Despite this, it still took the two of them almost ten hours to completely complete the development and inspection of the optimization program.

After all, in addition to optimizing the five independent variables of the three-section wing itself, Chang Haonan also added a flow deflector, making it extra difficult for himself.

"call--"

After checking the code one last time to confirm that it was correct, Xu Yang pressed the Enter key, submitted the calculation task, then leaned back in his chair and exhaled a long breath.

The next step is to wait for the results to come out, and then decide whether it is necessary to perform a separate optimization calculation on the design parameters of the flow deflector according to the situation.

"Is your canteen here... still open at night?" Chang Haonan asked hesitantly, looking at the moon that had risen into the sky outside the window.

From the intense work in the morning to now, he was really hungry.

"The cafeteria is closed, but I can treat you to Xiaozao." Xu Yang raised his wrist and glanced at the time, then replied.

"Do you still have a small stove here?"

"Of course, at least half of us are night owls, and the canteen is packed by midnight."

Xu Yang said, loosening his bun, and then arranging his hair slowly:

"It just so happens that I also have some questions about the differential evolution algorithm that I would like to ask you."

…

After some time, Chang Haonan, Xu Yang and Yang Xia sat around a table in the small cafeteria of the Institute of Computing Technology. The big fan next to them was blowing wind in this direction, barely breathing in the hot summer night. Brings a bit of coolness.

"I must admit that this morning, I still underestimated you too much, thinking that you just expanded the application scope of the differential evolution algorithm."

Xu Yang, who left the office area, put on a mid-sleeve floral shirt and was using chopsticks to stir the noodles in the bowl in front of him while asking Chang Haonan:

"But after reading the entire calculation program and explanation, I discovered that you have actually fundamentally expanded the theoretical basis of the differential evolution algorithm. Can you tell me what you think?"

Chang Haonan did not answer immediately, but glanced at Yang Xia next to him.

Since the Chinese Academy of Sciences asked Xu Yang to assist in its calculations, there must be no problem with the latter's confidentiality level.

But after all, places like the cafeteria were crowded, and he was not sure whether he could discuss the problem for a while.

Xu Yang, who was opposite him, naturally saw the reason for Chang Haonan's hesitation.

"Don't worry, I understand the rules. I will only ask about things related to algorithm theory, not your specific project."

At the same time, Yang Xia also nodded slightly, saying that if the discussion was only in the theoretical category, it would not be confidential, and the Institute of Computing Technology itself was also a confidential unit.

After confirming that there was no problem, Chang Haonan finally spoke:

"As one of the developers of the differential evolution algorithm, you should know that it mainly uses the differences between candidate solutions to generate new individuals, so it has huge advantages in global search speed and global search capabilities. My optimization process involves many An independent variable, it should be normal to think of this method, right?"

After hearing this answer, Xu Yang raised his head and glanced at Chang Haonan with a strange look. After a while, he spoke again:

"Do you want to test my theoretical skills, or do you simply want to hear me praise you?"

She didn't give Chang Haonan time to intervene again, but continued directly:

"The differential evolution algorithm was originally designed to solve continuous optimization problems, and it is an unconstrained optimization algorithm, but what you are facing is a combinatorial optimization problem with constraints."

"So I have been thinking today about what I would do if I were to face this problem."

"My final answer is to use "subtraction" and "addition" defined based on continuous nature to calculate the difference vector. However, although this idea is barely feasible, it lacks an operation method for the characteristics of the combinatorial optimization problem, and repairs must be introduced. Operations can ensure the feasibility of candidate solutions in the evolutionary process.”

"But your idea is completely different. You optimize or design three new differential optimization algorithms from the perspective of the algorithm itself."

"Today I roughly understood the first one, which is based on position rather than the traditional addition and subtraction operation based on numerical values. In the crossover stage, a crossover operation based on arrangement is introduced to avoid destroying the good structural blocks formed and ensure understanding. Converge quickly.”

"As for the other two, I can only understand them roughly. I even dare to say that even if you show your calculation program and design instructions to my teacher, he won't be able to fully understand them in a short time."

"How about it, are you satisfied with my compliment?"

Xu Yang put his chopsticks on the edge of the bowl and looked at Chang Haonan playfully:

"Besides, you just met me today. If someone else saw your code without a single comment, he would probably be furious."

"I……"

In the conscience of heaven and earth, Chang Haonan definitely did not have the two meanings that the other party thought just now.

He just simply didn’t know how far the intelligent algorithms had developed in 1996.

"It's really not convenient to explain this at the dinner table. To put it simply, my optimization of the differential evolution algorithm is divided into three stages, corresponding to three different types of engineering application scenarios."

Chang Haonan smoothly mixed a clove of peeled garlic into the noodles, and then continued to explain:

"The first is the first one. You have already seen some of the key points, but not all of them. To speed up the evaluation of solutions and further improve the optimization performance, I also designed a fast combination heuristic method based on the idle time increment matrix. Local search strategy..."

"As for the second algorithm, in order to increase the diversity of the PDE population, avoid unnecessary repeated searches, and jump out of local extreme values, the tabu search is combined with the permutation-based discrete differential evolution algorithm..."

"..."

Different from when he taught fluid dynamics simulation at Beijing-Hangzhou University, Chang Haonan did not originally understand the theory of this differential evolution algorithm, but just learned it through the system, and it was limited to the three types used in the wing optimization process. , so although it is not a problem to apply it, it is indeed difficult to explain it in a simple and profound way.

In the end, the two of them chatted until the staff in the canteen started cleaning and getting ready to get off work. Chang Haonan couldn't fully explain it, so he had to say that he would do it next time.

"In short, it is difficult to explain these things clearly in a short period of time. If you are interested, I can write you an explanatory document on differential algorithm optimization after I return to the guest house."

In fact, even if there was no such thing, he was going to go back and sort out the relevant content of the differential evolution algorithm carefully.

Although the system will only provide knowledge directly related to the project, Chang Haonan himself is not a complete novice. He can draw inferences from one example and deepen his understanding through his own learning ability.

Xu Yang on the opposite side seemed to still be digesting what he just heard from Chang Haonan, so he just nodded blankly, and after a while he said with some emotion:

"I didn't expect that in the past few years since I was studying abroad, my domestic colleagues would not only catch up with the world's first-class theoretical level, but also be ahead of them in terms of technology application. They are far ahead!"

"It seems that I made the right choice to return to China after graduation..."

Looking at Xu Yang whose thinking was gradually beginning to change, Chang Haonan thought for a moment and finally decided...

Let her be.

╮(╯▽╰)╭