What happened on the sea quickly reached Chen Nian's ears, but he was not surprised by such saucy operations.
Isn't it a basic operation of various countries to use warships that are originally meant to be decommissioned for fishery administration and coast guard?
It’s just an international practice, nothing to be proud of!
What's more, the 051 used for this "renovation" is already very old. According to his memory, it will be retired in 2006. Before retiring, it is nothing to use the remaining heat and temporarily save the situation.
However, for him, after this incident, the island building matter was put on the agenda, the obstacles were basically removed, and the remaining work was to build our own Sky Whale in the shortest possible time.
Chen Nian did not go to the scene to direct, because compared with all previous projects, the overall technical difficulty of the Tianjing was not as high as imagined.
Its difficulty is mainly reflected in the engineering difficulty.
In other words, the difficulty lies in how to dismantle and distribute a giant project in an orderly manner, how to mobilize the productivity of various departments, and how to carry out streamlined and standardized production.
Chen Nian actually couldn't help much in these aspects.
All he can do is to use his own strength to help the project team break through quickly when they encounter technical barriers.
Therefore, it doesn't make much difference whether it's on site or not.
Therefore, during this period, Chen Nian stayed at NPU to continue his studies.
He has completed all courses on advanced engineering materials, but unlike those written in fantasy, after learning a complete "secret book", he did not immediately obtain powerful "skills".
Because the accumulation of this thing is step by step, there is no chance of reaching the sky in one step.
However, looking back, this entire period of study provided him with more than 15 source points. Overall, it was extremely cost-effective.
Next, Chen Nian plans to continue to study material surface/interface science to further complete and enrich his knowledge system of materials science.
The course has been reserved, and the instructor is still Mo Zhongmin. With his level, he is quite comfortable teaching Chen Nian now.
In addition to attending classes, what Chen Nian does most currently is discuss drone system issues with Wang Jiancheng.
This was originally Wang Jiancheng's project, but after seeing the potential of the project, Chen Nian couldn't help but get involved.
It's not that I want to pick the fruit of Wang Jiancheng, but just out of a scientific researcher's curiosity and desire to conquer.
He wanted to know whether he had the ability to produce breakthrough results without the help of the system.
This kind of thinking is the same as that of high-ranking professional elites who continue to apply for jobs naked.
All to test their true strength and avoid boiling frogs in warm water and dying in the pot.
So far, the progress has surprised him.
Although Wang Jiancheng is indeed an out-and-out genius in this regard, his guidance and assistance also played an irreplaceable role.
After fully understanding the principles of the convolutional neural network algorithm, the two spent less than three weeks to create the first version of the weak artificial intelligence training model and were ready to start preliminary testing.
As an important project partner, Chen Nian will naturally not miss this significant test.
After a day of small classes, Chen Nian and Wang Jiancheng made an appointment in the laboratory of the school's Computer Comprehensive Education Center. There are minicomputers there that can support the computing power required by the convolutional neural network model, which can be used for this training. experiment.
——
Of course, the approval was actually the result of Chen Nian's secret arrangement through Chen Guo.
Otherwise, although Chen Nian could come and go at will, as an ordinary student Wang Jiancheng, no matter how awesome he was, this thing would not be open to him.
Chen Nian sometimes couldn't help but feel a little emotional.
Scientific research is like cultivating immortality, and both the wealthy and the Dharma are indispensable.
He walked all the way towards the laboratory. After entering, Wang Jiancheng was already waiting there. When he saw him walking in, he immediately stood up and greeted him warmly:
"Senior! Over here!"
Chen Nian went over and saw that not only Wang Jiancheng was present, but also Chu Jianshu, a professor at NPU who specialized in flight control, whom he had asked for help before.
At that time, Wang Jiancheng planned to start the project, but the resources and information preparations in all aspects were not sufficient, so Chen Nian contacted Chu Jianshu through Chen Guo's relationship and asked him to be an interface person.
It was also through his relationship that this laboratory equipped with a minicomputer was opened.
The other party was not a member of the Spark Team and did not know Chen Nian's identity. In his eyes, Chen Nian, like Wang Jiancheng, was an "ordinary" and relatively talented student.
He quickly went up and said hello.
"Professor Chu, are you here?"
Chu Jianshu smiled, nodded, and replied:
"This project of yours is quite interesting. I followed it before. Now that it has produced results, I must come and take a look. Doesn't it affect you?"
"It doesn't matter, it doesn't matter. If you are willing to provide guidance, it's not too late for us to be happy."
Chen Nian replied with a smile.
On the other side, Wang Jiancheng has set up the basic parameters of the model and connected it to the flight simulation program.
In subsequent training, the model will automatically start learning based on the random instructions input by the flight program, and use high-frequency trial and error to gradually find the strategy with the greatest benefits.
"So how did you set the criteria in the end? Is it still based on the following distance as mentioned before?"
Chu Jianshu asked.
"No, it has been changed. This criterion is too rigid. If you only use distance as the criterion, although the success rate is high, at the same time, if it fails, the consequences will be serious."
"We can't add estimated losses as a criterion, otherwise the entire model will become too large."
"Even if our small aircraft can run now, it won't be able to run on a drone in the future."
"So, the final strategy is to use the success rate of returning to the team as the primary criterion, and distance only as an auxiliary."
Hearing Chen Nian's answer, Chu Jianshu's eyes lit up.
This is indeed a very smart strategy.
As an important part of the formation, there is no doubt that UAVs must form a flight formation with manned aircraft. However, due to differences in performance, the formation of the formation cannot remain stable forever.
——
Not to mention mixed manned and unmanned formations, even a pure manned-machine formation cannot always maintain formation in battle.
But on the other hand, if you want to form combat effectiveness, formations and groups are necessary.
This requires pilots to quickly return to their positions after completing tactical maneuvers.
Using this as a standard to constrain the actions of drones and serve as a criterion for machine learning training, on the one hand, it gives the model a greater degree of freedom, and on the other hand, it can also meet the needs of actual use. It can be said to be a compromise. The best choice.
These two guys are indeed geniuses. They know one method and know all methods. For this model, what they master is definitely not just a single piece of artificial intelligence knowledge.
Object-oriented development is truly useful development.
However, how could an ordinary student have such knowledge?
——
He didn't know yet that Chen Nian at this time actually had practical experience in understanding the air force's tactics.
"Very smart, this should be our best solution. Shall we start now?"
"Start now!"
Wang Jiancheng nodded and tapped the Enter key with his finger.
As the fan of the minicomputer rotated, the program immediately started running.
Pieces of data are continuously input into the model. Since there is no visual interface, all learning processes are presented in the form of data.
Chu Jianshu couldn't quite understand what he saw, but Wang Jiancheng, on the other hand, stared at the crazy flashing lines of code and kept explaining:
"I have completed training a thousand times, and the effect is very poor. The success rate is 0%, but it is normal. I estimate that it will take at least a million training times to achieve the initial effect."
"Five thousand times, this is considered a success. A drone has returned to the team, but this is luck, and the strategy has not yet been formed."
Time passed by, and as the program continued to run, Wang Jiancheng's face became more and more solemn.
There is no other reason, the success rate is too low!
The training effect of this model was far weaker than he expected. If this efficiency was maintained, hundreds of millions of levels of training might not be enough.
What the hell is that?
There is an iron law in machine learning:
When the AI training intensity is greater than the manual training intensity, the model is invalid.
Obviously, this is the case with the model in his hand.
"It doesn't work. This algorithm doesn't seem to meet usable standards."
"The training efficiency is too low, and the speed of policy generation is also slow."
"If we continue like this, it would be more cost-effective for us to manually traverse the customized strategy."
Hearing his words, Chen Nian also frowned.
He asked:
"Can you see the reason?"
"I can't tell. I feel that there shouldn't be any big problem with our algorithm, but cold start consumes too many resources."
“The essence of machine learning is also traversal, but it cannot make any reference to perceptual experience.”
“In other words, artificial priors and machine priors are not aligned.”
Looking at Wang Jiancheng's slightly disappointed eyes, Chen Nian continued to ask:
"What about the confusion matrix we talked about last time? Can it be solved?"
"It can't be solved. Our problem now is that we have relatively few priorities and experience. If we want to obtain high-level solutions efficiently, we must require the machine to have an extremely high level of intelligence."
"And currently, the generalization of this task, that is, the difficulty of generalizing the task itself, is very difficult, which further increases the requirements for the machine."
"No, running away like this is not a problem."
With that said, Wang Jiancheng directly hit the stop button, and the model stopped running. The final result was that in more than 60,000 training sessions, effective results were achieved only twice.
This amount of data is simply not enough to support the continued training of the model.
Wang Jiancheng looked at the results on the screen with disappointment. At this moment, he felt an indescribable sense of frustration.
In fact, at the beginning, he was extremely confident in this model.
After receiving Chen Nian's advice and gradually understanding the basic principles of the convolutional neural network algorithm, he almost felt that the whole world was in his hands.
The potential of this set of algorithms is really great. As long as we continue to develop it, it is not an extravagant hope to one day completely change the world.
Moreover, on this track, I am already countless steps ahead of other researchers. As long as I can produce usable results and cases, I will most likely become one of the pioneers of the new generation of artificial intelligence together with my seniors.
What an honor this is, it goes without saying.
With such confidence, he could be said to be making great progress all the way, so fast that Chen Nian was amazed.
He originally thought that things would develop smoothly like this, but unexpectedly, the results of the first training of the model gave him a heavy blow.
It’s not just that the results don’t meet expectations, it’s that it’s not usable!
If it fails to meet expectations, you can move closer to the goal through continuous adjustments and optimizations, but if it is not available, then it is really useless no matter how hard you try.
Starting over might be the only way.
But can I still have such a spirit?
Chu Jianshu listened silently to the conversation between the two. In fact, this result was within his expectation.
You know, this is artificial intelligence!
Even if it is science fiction, the vision for artificial intelligence is still decades away.
In this era, if we want to achieve primary artificial intelligence under the current computer computing power, we must also consider the performance upper limit of the terminal. This is easier said than done?
Being able to do this is already a success in a certain sense.
The next road still needs to be walked slowly step by step.
——
Of course, when it comes to solutions, he actually doesn't have them.
Before, because of his interest in the project, he had already learned some knowledge about machine learning and convolutional neural networks.
According to the discussion just discussed by Chen Nian and Wang Jiancheng, the biggest shortcoming of the entire model is the lack of priori. If a large enough sample database can be implanted in the model and the machine can perform deduction and learning according to the samples in the database, then the machine training The difficulty should be reduced exponentially.
Such a strategy is not easy for others, but it is not difficult for him.
If you want data, it's just a matter of a few phone calls.
For its own scientific research purposes, wouldn't it be enough to ask the military to send over some training records of pilots? If there really isn't any, just let them fly a few times according to the target logic.
But the problem is that this method of solving the problem is tantamount to destroying the potential of the two children.
This time, I can help them solve the problem through my privileges, but what about next time?
If this is an urgent scientific research task, no matter what means are used, Chu Jianshu will not think it is excessive.
But now, it's more of an exploration.
Talents need to be cultivated. During their leveling process, you can give them some help and point the way, but you can never help them defeat the boss.
Thinking of this, Chu Jianshu asked calmly:
"Then what are you going to do now?"
The two were silent for a moment, and then Wang Jiancheng spoke:
"There are two ways to go."
"Or, I'll find a way to contact the aircraft model club and ask them to brush up the prior data for us."
"But the accuracy of the data produced in this way is relatively poor. I don't know how much time it will take for secondary processing."
"It's very possible that by the time the data is processed, Jiancheng will have missed the time to apply for the special class. But as long as the model can be written out, it won't be a loss."
"There is another way, which is to tear down the entire model and start over."
"I thought about it carefully just now. In fact, there are still many overlooked efficiency optimization points in the model. Once these points are solved one by one, there will definitely be benefits."
"The problem is that if we do this, the underlying logic will also need to be modified, and the amount of work will be relatively large."
Wang Jiancheng sighed helplessly, raised his head and looked at the ceiling.
"At least it's better than having no choice. If there is a problem, we have to find a way to solve it. We'll think about it when we get back."
Chen Nian nodded slightly and said nothing.
In fact, what Chu Jianshu can do, he can also do.
But his original intention of joining this project made him unwilling to use this somewhat "opportunistic" method.
This is a bit embarrassing.
However, besides this method, is there no other way?
He always felt that he seemed to have overlooked something. What was it?
On the other side, Chu Jianshu nodded happily after hearing Wang Jiancheng's answer.
Fortunately, my first reaction was not to give up, nor to subconsciously seek help from myself, but to plan a solution.
This shows that, at least in terms of character, they have basically passed the test.
So, he patted Chen Nian beside him on the shoulder comfortingly and said:
"It doesn't matter. That's what scientific research is like. Sometimes it takes hundreds or even thousands of failures before you can succeed once."
"Take your time, you still have time."
"Understood, Professor Chu. Don't worry, these are all minor problems. Let's talk to the teachers from the School of Mathematics tomorrow to see if we can do some optimization on the algorithm."
"That's good! Come on, let's go have a late-night snack? I'll treat you to it!"
After hearing his words, Chen Nian smiled slightly and stood up calmly, not forgetting to remind Wang Jiancheng to return the initial settings of the minicomputer.
After cleaning up, the three of them walked out of the school gate and had a bite of food together. During the process, both Wang Jiancheng and Chen Nian were a little silent.
Of course Chen Nian was thinking about the fleeting flash of inspiration that flashed through his mind just now. As for Wang Jiancheng, he was probably thinking about the model as well.
Seeing the performance of the two people, Chu Jianshu didn't say much.
He was still thinking that maybe the two children were shocked and didn't want to talk. Of course, they might be planning the next step, but no matter what, he shouldn't disturb them too much.
Things that need to be experienced will always have to be experienced. Since you have made up your mind not to interfere, you have to follow through.
However, when he returned to his residence, he still couldn't help but feel worried.
Many geniuses actually died due to one or several failures.
Especially young geniuses - because their minds are not yet mature at this time, a small setback may cause a huge butterfly effect, and ultimately cause irreversible effects.
Could Wang Jiancheng be such a person? Could Chen Nian be such a person?
Chu Jianshu hoped that they were not, but he was not too sure.
——
However, such worries lasted less than two hours.
Just after he finished washing, his personal mobile phone received a call from Wang Jiancheng.
On the other end of the phone, Wang Jiancheng said excitedly:
"Professor Chu, we want to borrow a minicomputer tonight! We know how to do it!"
"What? How?"
Chu Jianshu asked in surprise.
"That's it. Senior Chen Nian just told me that our thinking at the beginning was wrong."
"We have brought the human control of the aircraft into the drone!"
"For us, determining the target, then turning and moving is a continuous action, but this is not the case!"
"From a machine perspective, the so-called position changes are actually changes in data points one by one."
"We don't need to treat it as a continuous function at all. Instead, we should split it into two different models."
"The first model, dealing with location selection, is used to determine location targets."
"The second model will deal with policy execution."
"In this way, the complexity of continuous functions will be greatly reduced, and our training difficulty will be reduced by at least 90%!"
Chu Jianshu listened to his words silently. After a few seconds of silence, he said sincerely:
"You guys did a great job!"
At this time, an idea came to his mind:
Sure enough, in front of a true genius, there are really not many "setbacks".
(End of chapter)
