Chapter 135 Academic Pretense is the Most Refreshing (First Update)
There was a reason why Tong Yongshan's footsteps were heavier than usual.
Before returning to China, he was a distinguished professor at NUS.
Originally, he wanted NMU to support an associate professor specializing in deep learning to join the team, and he also compromised on NMU's request for co-first authorship.
As a result, something strange happened at XJTU, and they even wanted to be the corresponding author.
Finally, the new School of Statistics even announced that the foundation of this topic must be based on computer science. Computer science is theory, and econometric models are just applications. If Tong Yongshan disagrees, then they will engage in academic competition.
In the end, they naturally parted on bad terms.
If it really doesn't work, we can only discuss it with the National Institute of Planning.
But the level of the National School of Planning... Well, let's see if we can find gold in the shit.
Without experts in machine learning, this project would be impossible to carry out.
The theme of the group meeting had been emailed to everyone a week ago. Cheng Feng was the only one with the most experience in the group, so Tong Yongshan asked Cheng Feng to open the meeting as a sample.
The main content of this group meeting was literature reporting. Tong Yongshan had previously asked each person to carefully read three papers on big data and strong factors. The specific papers were chosen by the graduate students themselves.
Tong Yongshan quite likes Cheng Feng. Cheng Feng's personality is very similar to his, both of them are more focused on academics.
He was rather dull and socially anxious when he was young. Even after going to Miami and getting a tenured position, his social skills were only at the average level.
Cheng Feng often fails to express himself clearly when chatting with others, but when it comes to his profession, he is not timid at all.
These three papers are basically some extended research on strong factor models, time series and panel data published in A journals in the past year, which is also the area he has always paid attention to.
Cheng Feng spent nearly half an hour talking about these three papers from the aspects of research background, research results, conclusions and research significance, followed by some group discussions.
In fact, there was not much to discuss. Fang Yu, as a sophomore, was basically just an auditor. The other two first-year graduate students still hadn't figured out what a literature report was. They only started to get the hang of it after seeing Cheng Feng's report.
As for Jiang Nanzhen, the fake doctor, he was speechless. He lowered his head and wrote and drew in his notebook, looking very serious.
So the discussion basically turned into a question and answer between Cheng Feng and Tong Yongshan.
"Cheng Feng, you did a good job finding these three documents, and your analysis was quite accurate."
"But the main issue still lies with the two questions I raised with you earlier: First, relevance. Your three papers are not sufficiently correlated, and there is not enough connection between them."
"The second question is about the scope of your reading. The three papers you selected are not only not relevant enough, but also very narrow in their research direction. You can read some papers in related fields later, and don't just focus on factor models and panel data algorithms."
Cheng Feng nodded and ended his report.
"Zhang Chang, Song Liheng, who of you two will start first?" Tong Yongshan pointed at the names of the other two first-year graduate students.
One of these two people is from the School of Mathematics and the other is from the School of Computer Science. Zhang Chang from the School of Mathematics is a graduate from a national university who was admitted to graduate school, while Song Liheng was recommended for admission to the Computer Science major at USTC.
Both of them were attending a group meeting for the first time, but since they could become Tong Yongshan's graduate students, they were naturally not fakes. Although they lacked experience, they could make up for it with their abilities.
Zhang Chang, in particular, had a great tongue and was quick to react. Everyone could see that according to his PPT presentation, his opinion should be A. However, after listening to Cheng Feng's report, he immediately realized that there was a problem with his final conclusion, and that Tong Yongshan might not agree. In the end, he just used his tongue to twist his opinion into B.
After the two of them finished their reports, Jiang Nanzhen was the only one left.
The projector was connected to Jiangnanzhen's computer, and Jiangnanzhen turned to her own PPT with ease.
Wow, this is a gorgeous and elegant fresh look.
When I say it's elegant and fresh, I'm referring to this PPT template.
Most of the templates used by others are just found randomly on the Internet. Boys often don’t pay much attention to this. Being able to meet one alignment principle is already worthy of the tutor’s eyes.
As for Jiangnan Zhen's template, it's obvious that she spent money to have it customized by Taomao.
The visual style is extremely simplified, the chart style is consistent with the PPT style, the font is consistent with the rounded icon, and the illustrations are specially added with a desaturated color filter.
No, what kind of literature did you read? Why is there an illustration?
Speaking of a cluster of flowers, Jiang Nanzhen has simply mastered the PPT skills to perfection.
Fade, fly out, fly in, separate... Jiang Nanzhen's picture animations are all different.
As for the content, mia, mia, mia.
"Okay, no need to talk anymore." Tong Yongshan frowned. "What the hell are you talking about? Since when has Wikipedia been used as evidence?"
"What's the relationship between your third point and the first? What's the derivation process?"
"You said your previous research area was financial mathematics. I've also read your thesis. Did you write your graduation thesis yourself? How did your previous graduate advisor help you graduate with your master's degree?"
Tong Yongshan's words were already quite harsh, and Jiang Nanzhen's tears began to fall.
Everyone else held their breath, not daring to breathe.
Forget it, she is a girl after all.
Seeing Jiang Nanzhen's appearance, Tong Yongshan's heart softened.
"Be prepared before the next group meeting."
Tong Yongshan didn't want to go too far, frowned, and added a final sentence.
"Teacher, I'm sorry, I... I've been working for more than a year, and I'm not used to it... wuwuwu."
Although Jiang Nanzhen's appearance score is only 70 points, her appearance rose to 80 points after crying, which is quite magical.
Except for Zhang Chang, the other boys didn't have girlfriends, so they had never seen such a thing. I just felt that it was so pitiful that such a gentle person like the senior sister was criticized like this by the teacher.
Zhang Chang rolled his eyes, a slight look of disdain showing at the corner of his mouth.
Tong Yongshan waved his hand, ignoring Jiang Nanzhen who continued to sob, and pointed at Fang Yu: "Fang Yu, you should also speak. You are a sophomore, so you don't need to be too burdened, and you don't need to focus too much on performance. Just have content."
Fang Yu nodded, took the PPT remote control from Jiang Nanzhen's hand, and switched to his own part.
As the summarizer, Jiang Nanzhen had seen Fang Yu's PPT, but didn't quite understand it.
It’s not that Fang Yu’s PPT is so profound, it’s just that the content is too little.
Basically, each page has one or two words and one or two charts, nothing else.
And these one or two words seem vague and it is difficult to understand what they mean.
Jiang Nanzhen has seen this style of PPT many times in the year she has been working. There are several big scammers in the financial company where she works who are good at giving PPT presentations, and they basically all use this style.
However, this is Tong Yongshan’s group meeting!
You won’t pass the test by just fooling people!
Zhang Chang was able to force his way through because he had actual content. You don’t even have actual content, so just prepare to be scolded!
I was scolded for everything I did just now.
Jiang Nanzhen raised the corners of her eyes slightly, waiting to see the show.
"Hello, teacher, hello everyone. My name is Fang Yu, and I'm a sophomore majoring in finance."
"I have mainly read three papers in the past two weeks: 'Fast Learning Algorithms Based on Deep Belief Networks' published in Neural Computing in 2012, 'Autoencoding Variational Bayes' published at the ICLR conference last year, and 'Generative Adversarial Networks' also published at the NIPS conference last year."
None of the articles Fang Yu mentioned have anything to do with econometrics; they are basically all related to neural networks.
Jiang Nanzhen had just been criticized by Tong Yongshan, so it was not appropriate for her to say anything more. Cheng Feng was a dull person and would not express any opinion.
Zhang Chang and Song Liheng looked at each other and both thought it was nonsense.
puff!
Song Liheng almost couldn't help laughing out loud.
This little junior brother is a bit awkward in pretending to be cool.
He himself studied computer science, and as an outstanding graduate of the computer science major at USTC, his undergraduate thesis was on machine learning. Although his research was not in-depth, he still had some basic understanding of it.
But among these three papers, he only glanced over the paper "Generative Adversarial Networks" published by Ian of Google.
There's no way around it. That paper was so popular last year that I had to read it just to join in the fun.
But I hardly understood.
As for variational Bayes, it is not easy to understand.
A sophomore majoring in finance, reading this kind of paper?
Hahaha.
Song Liheng looked at Fang Yu with a look of sympathy and remembrance.
It reminded me of the time when I was in junior high school, holding a copy of "One Hundred Years of Solitude" in the classroom.
It reminded me of the time when I was in high school, holding a photocopy of the full-text book "Deep Understanding of Computer Systems" in the classroom.
I remembered the time when I was in college and I was loudly discussing with my roommate in front of the girl I had a crush on whether to go to Ali or Tencent after graduation.
Being young is so foolish.
It’s great to be young.
"The paper "Fast Learning Algorithm Based on Deep Belief Networks" introduces a fast learning algorithm for deep belief networks (DBNs), which is an important milestone in deep learning research. It shows how to pre-train deep networks layer by layer and then fine-tune them using backpropagation."
“This paper introduces deep belief networks, a multi-layer generative model composed of a stack of restricted Boltzmann machines…”
One of the core contributions of the paper is an efficient layer-by-layer greedy pre-training algorithm. The algorithm trains RBMs layer by layer…
"The paper presents experimental results on the MNIST dataset, demonstrating that DBNs perform better than shallow networks and other traditional methods."
…
A "snap" sound interrupted Fang Yu. Everyone looked at Jiang Nanzhen. Jiang Nanzhen quickly picked up a wooden hairpin that she had accidentally broken and smiled apologetically at everyone.
Just now she laughed in her heart at Fang Yu's PPTs that only had keywords and simple charts, but now every prompt word and chart was slapping her in the face.
Because these things actually work!
Is this kid really a genius?
How is that possible?
Of course, Jiang Nanzhen is not the only one who has this question.
Song Liheng stared at Fang Yu who was talking in amazement.
Are you a computer science major or am I a computer science major? I don’t understand either. Can an old kid understand what you are saying?
Not to mention that he didn't understand, Fang Yu himself didn't understand either.
But he knew how to recite the message transmitted to the Core of Etherand through the grapefruit.
It really feels best to pretend to be academic.
Fang Yu read the information in his mind and felt as if every pore was a poisonous dragon.
It feels so good.
“The second paper I studied was ‘Autoencoding Variational Bayes’, which is actually…”
“This approximation is achieved by maximizing a lower bound on the evidence, thus avoiding the intractable task of directly computing the posterior distribution…”
“The VAE loss function consists of two parts…”
"Basically, the VAE mentioned in this paper can be regarded as one of the underlying tools of machine learning, which can provide underlying support for the application of strong factor models in big data."
After listening to Fang Yu's explanation of the second paper, Zhang Chang could no longer sit still.
He is from the National Academy of Mathematics, and a considerable proportion of the content in this paper is related to his professional field, involving complex probabilistic graphical models and variational inference. Researchers need to have a relatively deep understanding of probability theory and statistics to be able to understand it.
And after Fang Yu's explanation, he actually understood what this paper was about!
This shows that this sophomore student really understood this article thoroughly.
Shouldn't this kind of person have gone to MIT or Stanford long ago? Or at least Princeton. Why is he still studying in China?!
Zhang Chang is a true academic genius with an IQ of 160. His IQ is not the kind of IQ that can be measured by a set of questions for 9.9 yuan on an online public account, but the result of actually taking the Mensa test.
If the path for him after graduating from graduate school had not been paved long ago, he would not have been able to stay in China for graduate studies.
No one understands what a god of learning is better than a god of learning.
This was the first time in Zhang Chang's life that he felt that his IQ was crushed by someone.
After Fang Yu finished the second chapter, he immediately started the third chapter.
"The third paper, 'Generative Adversarial Networks,' was a hot topic last year. The content is relatively simple..."
Hearing what Fang Yu said, Song Liheng felt his heart ache.
Is it simpler? Are you speaking human language?
How the hell is that simple?
Do you really understand the objective function?
“Where D(x) is the probability of the discriminator to identify the real data x, and G(z) is the fake data generated by the generator…”
Shit! I finally figured it out.
Are all undergraduates nowadays so perverted?
That’s not right. Before July, I was also an undergraduate student.
Under Song Liheng's lifeless gaze, Fang Yu began to conclude his group's first performance.
"At the application level, this paper mainly discusses the role of adversarial network training models in image generation."
"But in fact, based on its algorithm, we can try to apply generative adversarial networks to strong factor models."
"For example, in synthetic datasets, it can...; similarly, it can also be used for dimensionality reduction of high-dimensional data, with factor extraction efficiency hundreds of times higher than before; in abnormal data detection...; and most importantly, it demonstrates the potential to simulate and predict complex financial market behavior..."
"To sum up, I think the goal of the first phase of our project is to combine GANs with strong factor models. I believe that these two aspects should be able to produce brilliant sparks."
"My report is over."
Fang Yu nodded to Tong Yongshan and other team members and sat down naturally.
The scene was silent for more than ten seconds.
“Snap! Snap! Snap!”
Tong Yongshan broke the silence and was the first to applaud.
"Good! Good! Good!" Tong Yongshan's face turned red with excitement.
Fang Yu once again received Tong Yongshan's "Three Good Students" certification.
Tong Yongshan never imagined that in an ordinary group meeting, the undergraduate with the lowest academic qualifications in the group would find the direction for the first phase of his project.
"Fang Yu, if you were to combine GANs with strong cause models, how confident are you that you would succeed?"
When Tong Yongshan said this, everyone was shocked.
Has the team leader been appointed?
We have a bunch of masters and PhDs, and we have to listen to this undergraduate in the future?
Is this still an undergraduate?
Tong Yongshan's eyes were bright.
He doesn't care who is the team leader. No matter who is the team leader, they will work for him in the end.
He doesn't even mind whether he is the first author or the corresponding author. If Fang Yu can succeed, what's wrong with giving Fang Yu the first authorship of the first-stage paper?
This is the confidence of a top scholar with 7 papers published in the top five SSCI journals!
(End of this chapter)
Continue read on readnovelmtl.com
