Tech Hammer

What is it like to give a 60-minute report in a lecture hall with thousands of people at the world's top conference?

To put it bluntly, I really can't answer this question. If I had to answer it, I could probably only say one sentence, and I think it's okay.

Because even if you have already started to sit on the rostrum, you can clearly see the rows of heads below, and you can even find Professor Shen sitting in the first row, Ms. Lucy, who can switch styles freely, and the people he is most familiar with. Yu Xingwei.

But his complete thoughts were still not in the meeting. The mathematical formulas that kept popping up in his mind directly transformed the industry bosses and technical officers of the top technology companies below into mathematical symbols.

how to say?

The first time he felt this way, he started to struggle with the N-S equation and gave birth to the turbulence algorithm. The second time he felt this way, he began to study artificial intelligence and would rather apply for the postgraduate entrance examination a year later. EDA project; and today, this is the third time.

It's as if there is something in his brain guiding him in the direction he should go. When he encounters such in-depth explanations, it will activate and promote the secretion of dopamine, making him toss and turn, making him unable to stop...

Or maybe his brain wants to do something big, something big.

So nervous? nonexistent!

What mathematician would get nervous seeing thousands of mathematical symbols? Even though he was nervous, he was afraid that he would be tempted to apply for a new project again.

You must know that he promised Lu Dongyi that he would also go to the School of Mathematics of Yanbei University next year to continue his research on number theory.

At this time, Ning Wei was purely following inertia and expounding on the summary that he had already prepared. As for what questions these people would ask, he didn't even consider it. He didn't even bother to expand on the content on the PPT in the report, so as to extend the report time and shorten the question time.

So the report that was originally planned to be completed in forty minutes turned to the last page of the PPT in half an hour.

Of course, these dazed emotions were basically invisible to people in the audience, because Ning Wei's language was still fluent, and the necessary data references were readily available, with almost no lag.

So when he finished speaking so quickly, he was the only one surprised...

"Well, well, the above is an overview of the design ideas and methods of the turbulence algorithm. Thank you everyone."

There was a burst of applause from the audience, not too enthusiastic, but more polite.

After all, a groundbreaking paper that requires at least a week or more of research to have an outline can be condensed into an abstract that can be reported in an hour and a half. In fact, there cannot be too much intrinsic content.

At least he cannot describe in detail the processing of each data iteration in the paper, the divergence and convergence of functions, the meaning of each recursive determination, screening and associated functions or equations.

Ning Wei didn't know how many of the thousands of people in the audience had carefully studied his paper, but what he was sure of was that if he didn't do enough homework in advance, read his paper three times and take notes. In this case, you definitely can’t understand it.

So the applause didn't move him.

He just sat quietly in his seat, allowing himself to focus more on the meeting, waiting for the big guys in the audience to ask questions.

Similar to what he had thought before, the two big guys who first asked the question focused on the idea of ​​space transformation during algorithm iteration. This is also the most obscure part of the turbulence algorithm in a theoretical sense.

Therefore, Ning Wei was well prepared. It took him ten minutes to roughly explain the general idea and show the more detailed derivation process, especially the process of limiting high-order terms by the defined limiter as the flow field parameter space changes.

Then he saw on the stage that the microphone was handed to Lucy Rowan, the Harvard genius girl Feng Shaojie said.

"Hello, Mr. Ning, after listening to your report, I noticed that the latest experimental data you published in today's report are quite different from the experimental data in the original paper. It should be that there has been great progress in the latest testing stage. . In the process of studying the paper, I roughly understood that the algorithm learning stage mainly relies on the training set, and the training set comes from high-quality measured data rather than data simulation."

"So when I was researching the paper, I suspected that the reason why the laboratory error reporting rate described in the paper could not be reduced at a high level for a very long time was probably due to the noise interference caused by random errors. This is the most difficult problem to solve. place. But in your report just now, I found that the false positive error rate of the algorithm has been greatly reduced, so I would like to ask how this is achieved?"

Ning Wei looked at the solemn-looking girl in the audience in surprise, and roughly understood why Feng Shaojie said she was Harvard's darling. It was obviously not because of her figure, but more because of her mind.

He has never been in the laboratory for a day, but he can actually determine the reason for the high error rate of the turbulence algorithm before the update from his paper alone. This makes Ning Wei no longer dare to underestimate the world's heroes. This is An outstanding academic who has both a proud figure and a smart mind.

After all, they spent nearly a month observing data in the laboratory every day and still couldn't find the problem.

Just to answer this question…

"The method to solve this problem is not universal. In fact, it is solved by deepening the understanding of the NS equation itself. Simply put, it is solved by setting the high-order space to limit the data overflow. If you must listen to the complete solution, I I don’t know...if there is enough time." Ning Wei explained.

In fact, what he wanted to say was that he didn't know whether the girl could understand it, because the content in the field of pure mathematics was not within the scope of special research by algorithm engineers. But saying this was too hateful, so Ning Wei changed his tune temporarily.

He really wasn’t looking down on people, he just recalled the days when he and Lu Dongyi discussed and argued for five days...

However, Lucy in the audience just smiled slightly and said relaxedly: "There are still twenty minutes left. I believe that with your ability, you can answer everyone's questions."

Ning Wei glanced at Professor Shen in the audience and just smiled, then nodded helplessly and said, "Okay, then I need an electronic pen, thank you."

After saying that, Ning Wei directly removed the prepared PPT, created a new blank document, then took the electronic pen sent by the field staff and started writing on the document.

Soon a series of formulas were written and displayed simultaneously on the big screen behind them. Ning Wei also formally began to explain: "Can you understand this measurable function? I won't explain it in detail due to limited time. First, let's put all the formulas that meet the requirements. The minimum value of the above function conditions is recorded as ∥f∥Lp(·)..."

Why is there a saying in academic circles that the field of pure mathematics stands at the pinnacle of the contempt chain in academic research?

Because number theory is really highbrow.

To put it bluntly, number theory researchers and ordinary people view the world in a different way.

Especially after modern number theory has subdivided its fields, it is not accessible to ordinary people. Because many times those who study number theory at the same time may not be able to understand the thinking of their peers once they cross boundaries.

For algorithm engineers, mathematics is naturally extremely important, but after all, the research of algorithms is to solve the real problems we are facing now. For the research of number theory, it is not important whether it can solve real problems, or even whether it can solve problems that may arise in the future. It doesn't matter, the fun of its research is more towards the philosophical discussion of the nature of the world.

Take the study of NS equations as an example. Without NS equations, wouldn’t physicists be able to study fluid mechanics? The answer is obviously no.

The essence of physics lies in observation and experimentation. Experiments in different environments can be conducted not only hundreds of times at a time, but not even hundreds of times, or even millions of times in different environments. Through observation and recording, it is enough to summarize various general laws under different fluid conditions, and then deduce Provide various formulas for different situations.

In fact, modern physics does the same thing.

Of course, this does not mean that number theory is completely without practical significance.

At least the various models established through number theory can at least reduce the workload of physicists.

The most important thing is the process of solving problems and the mathematical methods used, which can often guide the development of scientific research to a higher level.

Similar propositions such as Goldbach's conjecture.

So when Ning Wei turned the question to the field of number theory, which he is currently best at, at least 90% of the people in the audience started to get confused...

It even reminds many people of their fear of being dominated by number theory.

This is a Computer Algorithm Conference. Are you here to talk to everyone about math problems?

…

Yes, Ningwei is talking about the topic.

At first, he would ask if everyone understood the previous sentence during a certain space transition stage. However, after he was immersed in the fun of review and Lu Dongyi's thinking storm, he was too lazy to care whether the people in the audience understood what he said. Something to say.

He was just answering questions at an academic conference. He was not giving a lecture. He really didn't need to care whether anyone didn't understand.

In fact, when he started to explain from the direction of number theory, it was already destined that this report would go in a strange direction.

The big guys in the audience should stay calm and keep their expressions straight. After all, if they show cowardice, they will appear to have been intellectually suppressed by a young man, and they will lose face...

But there were already whispers of discussion in the back row.

At the beginning, I could still hear people consulting and discussing with people around them: "What is the connection between this step and the next step? How to derive it?"

Someone else can answer: "If what I learned before was not completely handed over to the teacher, it should be Young's inequality."

But when Ning Wei began to discuss the problem of boundedness of operators in frequency space, there was no sound of discussion, and countless people started complaining mode.

"Brother, do you understand? What did he write?"

"No, I'm reflecting on why I'm sitting here. Can you tell me? Why are we sitting here?"

"Farke, is this a computer conference or a mathematician conference? Did the guy on stage go to the wrong set?"

"No, maybe we traveled to Princeton, you know, there are the most crazy people there!"

"I'm here to attend the Computer Algorithm Conference, not to be remade!"

"Calm down, it's normal if you can't understand! People who can understand shouldn't be sitting here."

"I finally understand why he has never seen a woman wearing a backless evening gown! He doesn't deserve it! Really, this kind of person only deserves to live with his right-hand man for the rest of his life!"

"What kind of joke is this?"

"Don't you know yet? The story of that Chinese boy who doesn't understand the charm?"

…

As for Lucy, the initiator of this situation, she was completely stunned. In fact, she was not much better than everyone in the back row. The process that she could follow and understand was probably limited to the first three steps of Ning Wei's derivation.

When more mathematical tools and theories were proposed, she began to feel confused.

She probably didn't expect that even if she was beaten to death, what she thought was an extremely difficult problem would actually lead to a real blow to the dimensionality reduction and what seemed to be a crushing of her IQ.

She tried very hard to digest what Ning Wei wrote and explained on the blackboard, but it seemed to be of no use. At this time, the content of the conversation between the instructor who was also the convener of the meeting and Professor Shen on the other side came intermittently to her ears.

"Shen, maybe we should invite a few mathematician friends in the field of pure numbers to sit in future conferences... You know... there are probably not many people here who can comment on his speech... This will be very embarrassing... Or maybe we He should be sent to Princeton to give a report..."

…

"In summary, at this point, it should be clear to everyone that we have effectively achieved global convergence in this way, and then brought this result into the algorithm, thereby effectively eliminating the interference of adjacent errors and allowing the algorithm to The learning process isolates noise interference from adjacent data and improves the accuracy of recognition."

"I don't know if you can fully understand this, but I think you should know that after all, I am still an undergraduate, and my current main learning direction is still in basic number theory. The design of turbulence algorithms is only the process of deriving NS equations. , the extension of new mathematical methods and tools. So it is really difficult for me to elaborate on this issue from the perspective of purely designing algorithms."

"Of course, it doesn't matter if you don't understand. I have written a paper on this topic and submitted it to the Journal of the American Mathematical Society for publication. You can refer to it for understanding. Although the review has not yet been completed, But I am confident that I can pass it, at least it helped me solve practical problems.”

"In addition, I have also put these data tools into a relatively alternative EDA software project that I am currently researching. When this project is completed, I hope to have the opportunity to communicate with you. Thank you."

After making his final concluding remarks, Ning Wei looked at the time. It was exactly sixty minutes. The time was perfectly controlled and he didn't even let the venue management rush him.

The only thing that embarrassed him probably was that the audience did not immediately burst into applause like when the speaker took the curtain call, which was probably out of politeness and respect, but still warm applause.

Instead, there was silence.

Perhaps what everyone is thinking at this moment is just that...it's finally over?

It wasn't until he stood up from his seat that someone in the front row finally started to applaud, sparsely, but eventually more people responded, and the applause began to warm. Well, of course, the vast majority of people probably applauded while talking. I scolded my mother in my heart...

"Why is it difficult to explain the problem from the perspective of designing algorithms?! Who are you looking down on when you say this?!"