Chapter 89 I really don’t understand artificial intelligence!
It makes sense for Zhang Zhiqiang to criticize Patrick Roche. There are countless scholars in the world who would criticize Patrick Roche.
Because, lucky.
Lucky people who win the lottery will also be sour, but winning the lottery often happens. There are many lottery winners around the world, and in recent years, only Patrick Roche has discovered the latest Mersenne prime number.
The name has been listed on the GIMPS official website for several years.
The GIMPS official website has an introduction to the luck of Patrick Roche, which reads, “For Patrick Roche, this season has been even more fortunate, and his persistence has paid off handsomely. For many years, Patrick has been using GIMPS software as his
A computer version of a free "stress test," which he started looking for on his own media server less than four months ago as a way to contribute back to the project."
"Some GIMPS participants have been searching for more than 20 years and tens of thousands of attempts without success. This proves that, with a little luck, anyone can find the next new Mersenne prime."
Prime software has hundreds of thousands of people online for a long time and millions of CPUs running it. Within a few years, only one Patrick Roche appeared. You can imagine how lucky he is.
In addition, discovering a new Mersenne prime number will result in a reward of more than 100,000 US dollars, and a huge increase in personal fame.
For example, the GIMPS website has long had Patrick Roche’s personal ‘lucky introduction’, probably to attract more people to join.
Many reports on the discovery of new Mersenne primes will mention Patrick Roche. His name appears in the pages of major media. People who don't know it think he is some famous scholar and a scientist who has made great contributions.
But in fact, Patrick Roche is only at the bottom of the academic circle. At most, he can be regarded as an algorithm engineer, and his ability is not too outstanding.
If you really want to calculate it, Patrick Roche is not a scholar, but because of his luck in discovering Mersenne prime numbers, he has become more famous than other real scholars with "achievement similar to academic achievements".
This is the source of envy and jealousy among scholars.
What did Patrick Roche do?
He just used Prime software to do server stress testing, hung up the software on his own computer and ran it, and then he was lucky enough to find the Mersenne prime number.
He gained money, gained fame, and left his name in the historical record of Mersenne prime numbers.
Wang Hao felt a little emotional when he thought about it carefully. He had to admit that Patrick Roche was indeed very lucky.
He shook his head, customized the program running time, and was soon assigned a very large prime number. He also started to hang up the Prime software on his computer, and then ignored it and began to study the Mersenne prime numbers.
He studies algorithms.
The main function of Prime software is to perform distributed processing and calculation statistics. In fact, the internal testing algorithm for Mersenne prime numbers is very simple. For verification of a number, the Lucas-Lehmer test is used.
The Lucas-Lehmer test is a primality test for Mersenne numbers. It was formulated by Edward Lucas in 1878. Fifty years later, Derek Henry Lehmer improved the method.
This primality test method for Mersenne primes has been very simplified.
To use the Lucas-Lehmer test to verify whether the Mersenne number 'Mn=2 to the nth power -1' is a prime number, you only need to let the program loop and run 'n-2 times'.
Of course, since extremely large numbers with tens of millions of digits are involved, the amount of calculation for each cycle is still very high, but the primality test for extremely large numbers with tens of millions of digits can be said to have been simplified to the extreme.
For example, an ordinary home computer can detect the primality of a ten-million-digit Mersenne number after running it for about a hundred hours.
This speed is quite amazing.
Because of this, for almost a hundred years now, the Lucas-Lehmer test is still used to test the primality of Mersenne numbers.
The content of the Lucas-Lehmer test can be understood at a glance, and there is nothing difficult about the method.
Just because the method is too simple, it seems impossible to start improving it.
For example, the method of ordinary addition is already simple enough, how can it be improved?
Wang Hao thought about it carefully and found that there was no way to start. However, the research must have a direction. The testing method for a single prime number could not break through, but it could also make a breakthrough in testing digital screening.
However, after thinking carefully for a long time, he still couldn't find the direction. He could only slowly see if he could get any inspiration.
…
The next morning, Wang Hao got up early and went for a morning run.
The rhythm of life has returned to its previous routine, and I have met familiar people when exercising.
Su Yingxue.
She still wore that sporty look and still looked dashing.
Wang Hao and Su Yingxue ran around together for two times and exchanged a few words with each other before they realized that the other party was planning to stay in Xihai City for a long time.
"I am responsible for following up on the NanoLab project." Su Yingxue gave a simple explanation, which revealed a lot of information. She advocated cooperation with NanoLab and was directly responsible for the follow-up of the project. She must have attached great importance to the project.
If there is no progress in the NanoLab project for a long time, it may have some impact on Su Yingxue's personal career.
"I remember saying that I suggested that you cooperate with Navitas Labs." Wang Hao smiled, "I will definitely not regret it."
"I believe you." Su Yingxue looked into Wang Hao's eyes and nodded seriously.
Wang Hao exchanged a few words with Su Yingxue and continued to run two more laps. He felt a little tired and left to have breakfast.
Soy milk, fried dough sticks, and a tea egg are a classic combination, and you can eat them happily to replenish your energy.
The day begins again.
A girl wearing a sweater and smiling sweetly came over. She placed the plate opposite Wang Hao and said with some surprise, "Teacher Wang Hao, I looked just like you. You really did, don't bother me."
Bar?"
"It doesn't matter."
Wang Hao raised his head and said something, and found that the other person seemed familiar. He was the counselor in Luo Dayong's class. His name was...
It does not matter!
The visitor is Meng Qian.
She has a cheerful and talkative personality, and she took the initiative to talk to Wang Hao after sitting across from him, so that the atmosphere would not be too dull.
"Teacher Wang Hao, congratulations to you. You have created a domestic scholar. In the past twenty years, you have accomplished international achievements in the field of analytic number theory, and you have also been in the news. You are really amazing."
"My friends and I were talking about you yesterday. You are now a big celebrity in the school!"
Meng Qian complimented Wang Hao several times in succession, and then talked about her own research, "Teacher Wang Hao, I remember you said last time that if you have time, please give me some advice. My thesis and doctoral thesis are too difficult. I'm here."
I can’t find my direction now.”
Wang Hao asked, "You are doing computer research, right? What field?"
"Artificial intelligence, algorithm." Meng Qian introduced it in two words.
Wang Hao immediately shook his head, "I can't help with this. I don't know much about artificial intelligence, but there are many people engaged in research in this area in the school. Zhang Zhiqiang in my office has done relevant research, how about I help you
introduce?"
"That's not necessary."
Meng Qian hurriedly waved his hand and said, "My tutor also specializes in research in this area, and you can't always ask others to do research. I am working on algorithms that represent learning, and I just want to find a direction."
"Algorithms that represent learning..."
Wang Hao thought and shook his head, "I really don't understand this field at all."
Meng Qian smiled and said, "Don't be humble, Teacher Wang Hao, you are an algorithm expert. If you have time, can you help me? I don't need to tell you what to do in detail. I just want to find a direction."
"Um...Okay." Wang Hao hesitated for a moment. It would not be good to always refuse. He thought it would be okay to go and see it, so he agreed.
But he still stressed, "I really don't understand the field of artificial intelligence. I won't be able to help you much when the time comes, so don't blame me."
"Of course not. I would be very grateful if you could help me take a look."
Meng Qian smiled cheerfully, and made an appointment with Wang Hao, and simply decided to set it in the afternoon.
When they were about to leave after breakfast, Meng Qian suddenly said, "By the way, Teacher Wang Hao, my name is Meng Qian. You can just call me 'Xiao Meng'. Everyone else calls me that."
She said and winked at Wang Hao.
Wang Hao suddenly felt a little embarrassed.
It doesn't matter if you forget the other person's name, but the other person still notices it?
Meng Qian was quite cheerful, and even asked herself to call her "Xiao Meng". The problem was, the other party should be two years older than herself, right?
Is it appropriate to call me ‘Xiao Meng’? How about changing it to ‘Sister Meng’?
…
When Wang Hao waited until the afternoon to see Meng Qian again, he still hadn't decided on the issue of addressing him, so he finally decided to call her "Meng Qian". It was definitely right to call her by her name.
"Meng Qian, where are we going?" Wang Hao followed Meng Qian and felt like he was about to leave the school gate. He could see a hotel sign in the distance and had a bad feeling in his heart.
Fortunately, Meng Qian turned a corner and entered the sixth-floor laboratory building.
Computer laboratory building.
The two floors above are the computer laboratory where Zhang Zhiqiang is located.
Wang Hao and Meng Qian went upstairs together, and met Chen Qinghua on the fifth floor. The little old man was walking with his hands folded behind his back. When he saw Wang Hao, his eyes suddenly lit up, "Wang Hao? You are a rare visitor!"
The little old man came over enthusiastically and pulled Wang Hao inside. "Let me show you around. Our laboratory has purchased two new servers. Their computing power is very..."
Wang Hao quickly emphasized, "Dean Chen, I'm just here to show Meng Qian something."
"Meng Qian?"
Chen Qinghua was confused when he heard this, and then he noticed Meng Qian next to him. He took another careful look at Wang Hao, suddenly realized and said in a long tone, "Oh~~~"
"You talk about yours, you talk, you talk..." He said and walked away with his hands behind his back, and sighed, "Young man!"
Wang Hao was a little embarrassed, and Meng Qian was also a little blushing, but it was nothing, and they went to a laboratory together.
There seems to be no difference between the computer lab and the office, except that the computers look more professional, and there are a lot of computer accessories piled in the corner, making the room look a bit messy.
There were several other people in the laboratory. When they saw Meng Qian walking in with someone, they didn't recognize him at first, but they quickly realized and gathered around and started chattering, "Teacher Wang Hao!"
"Professor Wang!"
"I studied your effective and irrelevant carry algorithm. It's really... too difficult. It's as difficult as the sky. After studying it for two days, my hair almost fell out."
"Teacher Wang Hao, do you have time to show me the research? I encountered a problem..."
Meng Qian quickly announced 'Wang Hao's ownership', "You all get out of the way, I want Teacher Wang Hao to give me some advice."
"well……"
Several people stopped talking.
Meng Qian opened the computer. There were several files in it. She clicked on one of them, which contained some code and the like. She also began to explain, "I am doing research on algorithms for representation learning."
"The basis is CNN (Convolutional Neural Network). Related research based on CNN is a hot topic, and this algorithm is always being improved."
"I also know about another machine learning algorithm GBDT..."
Meng Qian talked a lot in succession, and it sounded very professional. He must have put a lot of effort into it. Several other people in the laboratory are also doctoral students. Their research is not in the same direction, but they all have it.
Similarities and repetitions.
Wang Hao pursed his lips and listened carefully for a long time. When he was about to speak, Chen Qinghua and another professor walked in.
"Teacher Sun!"
"Professor Chen!"
Several doctoral students shouted politely.
Chen Qinghua also introduced Wang Hao, "This is Professor Sun Guangyuan."
Sun Guangyuan is a doctoral supervisor, and two of his students in the laboratory are his students. After shaking hands with Wang Hao, he immediately said, "Professor Wang, I'm sorry to bother you. Please leave us alone and continue talking about yours."
Chen Qinghua and Sun Guangyuan stood nearby and watched.
Two computer science scholars stood nearby, most likely engaged in research in the field of artificial intelligence. Wang Hao felt that he was giving advice to an expert from an outsider, so he quickly emphasized, "I am a layman, and what I said is wrong. Please don't
Joke. I really don’t understand artificial intelligence.”
Sun Guangyuan immediately laughed and said, "Professor Wang, what's the joke? Crossing fields is like a mountain. That's what research is like. It's hard to make progress if you just do research by yourself. Maybe you can think of ideas after listening to a few words from other people.
"
Chen Qinghua also said, "It doesn't matter whether what you said is correct or not, this is not a professional academic conference."
Wang Hao suddenly felt confident, and he said, "My understanding of 'representation learning' is that it is a core research issue in the field of artificial intelligence. Some successful models." Successful models can be understood as special cases of representation learning.
…”
After getting a start, a lot of knowledge poured into his mind. Combined with his understanding of the algorithm and his ideas, Wang Hao suddenly gained confidence and continued, "Use probability to model the hidden variables and the joint distribution of the observed variables.
Variable models, as well as deep models that learn hierarchical representations end-to-end, are both a type of representation learning.”
"We are now in the era of big data, and the main difficulty in research on representation learning is the analysis of extremely large amounts of data."
"Whether it is the noise of random algorithms, the instability of approximate optimization algorithms, or high time complexity, they all affect the efficiency of representation learning algorithms..."
As he spoke, he got to the point. He simply walked to the small whiteboard next to him and drew a simple model diagram before continuing, "Let's look at this diagram. In fact, no matter how it indicates the direction of learning or what kind of research we do, it is inseparable.
Open this diagram. Either analyze and simplify the algorithm, or start from the structure. The easiest and easiest way to produce results is to simplify the structure and form a direct path in a certain direction."
"If we walk from point A to point B and then to point C, it's definitely not as good as building a path between A and C."
"So, how to do research and how to build pathways?"
"I think there are two methods, one is to adopt a new structure, the other is to adopt a new analysis method..."
"Wow wow wow..."
Wang Hao spoke for nearly thirty minutes at a stretch, and a discerning doctoral student even poured a glass of water over to him.
Then, continue.
Finally, after about forty minutes, he ended his conversation.
Looking at the people in the laboratory who were deep in thought and understanding seriously, they immediately nodded with satisfaction and said, "I just have a superficial understanding, so I apologize if I say anything wrong."
"I am just a layman when it comes to artificial intelligence."
A group of doctoral students, including Meng Qian, immediately looked at him with admiration. They all felt that they had gained a lot after listening to a lot of explanations, and their research ideas became clear.
Sun Guangyuan and Chen Qinghua tugged at the corners of their mouths at the same time.