"Oops, it's so late! I'm sorry, senior, I don't have anyone around me to talk about suspense novels with me. I couldn't help but talk too much when I met a friend today. I'm really sorry for delaying your game."
"It's okay. You can watch the video anytime. It's hard to find friends. I had a lot of fun chatting with you. I want to thank you."
"Hahaha [handshake]. Let's end our conversation today. It's not too early, so I won't disturb the seniors."
"No bother, no bother. Good night then."
"Good night [moon]"
Ling Zhi put down his phone and was satisfied. But then he remembered that he felt a little guilty for Brother Feng, so he logged into WeChat on his computer and downloaded the data sent to him by Chen Feng to see how much it was.
At this time, Chen Feng and Gao Feng happened to return to the dormitory together.
"Lingzi, have you seen the data I sent you? Do you think it's good to screen?"
The so-called sifting of data actually means to delete pictures with unobvious features. For example, some eyeball lesions are densely covered with small dots, which are obvious features. Some of them are vague, but they are not obvious features, which is not conducive to the training process of the model.
Need to be deleted.
"Well, I'm still reading. By the way, you sent me 2,000 pictures at once. This is too hard."
Ling Zhi tried to change the topic to prevent Chen Feng from discovering that he hadn't actually screened yet. In fact, Ling Zhi knew that Chen Feng was already working very hard. He had tens of thousands of pictures, but only gave himself 2,000. His own workload was only
is the heaviest.
Chen Feng didn't notice anything and replied mindlessly:
"Oh, it's okay. You can sift slowly and don't rush. Sift as much as you want. Just give me the sifted ones when the time comes."
"Okay, I understand."
Ling Zhi breathed a sigh of relief, but secretly made up his mind to help Chen Feng finish the screening.
Before seeing Chen Feng process the data set, Ling Zhi always felt that this was a thankless job that should consume a lot of energy. In fact, when he called him to eat together one time.
"I'll go, you're so stupid at deleting data."
Chen Feng used the mouse to select a row of pictures in succession and deleted them without any hesitation.
"It's so obvious. I've looked at so many pictures, and I feel like I've eliminated them all if the vision isn't very clear, so as not to torture myself."
That's not to say that it's not good for him to do this, but Ling Zhi knows that if it were him, he would definitely look at the pictures one by one. He is not sure whether to delete some pictures, and may have to look back and forth many times before making a decision.
.
But having said that, the huge data base determines that even if he does this, it will inevitably consume a lot of time and energy. But even so, Lingzhi does not intend to change his style of screening data.
Ling Zhi went to take a shower first. After taking a shower, he discovered that Lin Qisong had also come back. The four people's day of scientific research life came to an end temporarily.
"Gaofeng, how was your experience last time when you accompanied Lao Wu to teach undergraduates? I haven't heard from you since."
Lin Qisong asked.
"Oh, that, don't mention it, the day of class was a car accident scene, you know."
Gao Feng stood up and said to Lin Qisong. Chen Feng and I were also very interested and sat in our seats and listened.
“He was not prepared for class that day. I couldn’t quite understand what he was saying on the ppt, and sometimes I didn’t know what to say. It was expected that there would be a lot of undergraduates coming to listen to his lectures, and most of them were students who wanted to pursue postgraduate studies.
As a result, half of them didn’t come that day for some reason, and two left in the middle. It made me want to laugh.”
"Wow! Wasn't it embarrassing for Old Wu the whole time?"
"That's not true. He tells his story, and he doesn't care whether the students below him listen or not."
"Hey, why did we choose him as our mentor?"
"I don't know either. Who told me to promote Lao Wu's great achievements to me at that time?"
"I didn't know it at the time, and I didn't ask the seniors in the past. I saw that the introduction on his official website was good, so I recommended it to you."
"Oh, we can only accept our fate. Undergraduate students are not like us. It is very costly for us to change tutors. Undergraduate students are very casual and will not come if they don't want to come."
Ling Zhi also feels the same way. Choosing the right mentor is crucial to his graduate career. As for his mentor Wang Haibo, Ling Zhi is still very satisfied.
Lao Wang is not the kind of "sheep-herding" tutor mentioned in some online tutor selection posts. The so-called "sheep-herding" means not caring about the students and letting them fend for themselves. Such a tutor is obviously irresponsible.
If we were to define a personality for Lao Wang's way of managing students, Ling Zhi felt that "semi-freedom" would be more appropriate.
Lao Wang will not interfere too much with your own research ideas and experimental details, but he will grasp the accuracy of your research ideas and the speed of your research progress from a broad perspective. He is obviously a responsible mentor.
To sum up, he usually doesn't care about you, but you want him to know what you are studying.
Ling Zhi felt that he was obviously lucky. In the past year of getting along with him, Ling Zhi felt that Director Wang was affable. Although he would inevitably act like an elder sometimes, he was still approachable most of the time.
Thinking about the rumors circulating on the Internet about those masters and Ph.D. students who were forced to fail in their studies by some harsh and irresponsible mentors, and some even committed suicide. Ling Zhi felt that he was very lucky, even though he had two choices when he chose Director Wang.
His eyes went dark.
"So you see, how stressful it is to sit in a small laboratory with Lao Wu. Especially Du Yao, who is sitting next to him, has a sad face all day long."
Gao Feng said.
Lin Qisong shook his head helplessly, turned around and turned on his computer and started watching his documentary.
A night of silence.
Early the next morning, it was the weekly group meeting of Lingzhi's laboratory.
Ling Zhi came to the laboratory early, set up the projector in the conference room next to it, and then copied the ppt he made into it. Today he will share the paper.
After a while, the students from the division came to the laboratory one after another, and soon after, the instructor Wang Haibo also arrived.
Wang Haibo, Ling Zhi’s mentor, is in his early 40s. He was just promoted to professor this year, which is a time of great satisfaction. I heard that his youngest son has just been born, which is also a time of double happiness.
Although he has been deeply involved in the computer field for many years, he has not worked too hard. The flat hair on his head is still lush. When he walks on the road, his steps are always windy, and he always smiles when he sees everyone.
Even as he got older, his cervical vertebrae often felt uncomfortable. Ling Zhi wanted to give Director Wang a massage device to try out on Teacher's Day, as a way to repay Director Wang's kindness in educating him.
“Does anyone have anything to share today?”
Ling Zhi said without hesitation:
"Teacher, let me do it."
Lingzhi opened his ppt and preached without fear of the situation:
"Today I would like to share with you my recent experimental results and papers."
…
"This is my data preprocessing process. I roughly divided it into 5 steps..."
"This is the result of my experiment on sentence pair matching. If two sentences are posted by the same person, they are positive samples. If they are not the same person, they are negative samples. ..."
"Wait a minute, what are the features you use for classification?"
Lao Wang asked.
"Oh, I will go into details later, I am just throwing out the results first..."
"Are the terms positive samples and negative samples reasonable? What do you think of the results of this experiment?"
Lao Wang discussed it with everyone for a while, and then asked Ling Zhi to continue.
“The feature I use is a 14-dimensional vector, including manually extracted features and features extracted by neural networks.…”
Lingzhi's presentation of ppt follows Lao Wang's previous teachings to his students - use more diagrams and less text. Lingzhi deeply agrees with this.
In fact, ppt is originally used to highlight the key points. If too much text is piled on it, the speaker will easily read to the ppt, and the listeners will find it boring and will not think for themselves. Using pictures and a small amount of text to highlight the key points will
It is easier for the audience to accept it.
Just like the paper shared by Lingzhi next, a picture is used to illustrate the distance relationship between each word in a sentence.
"For example, there are two sentences now, 'He spoke to the media' and 'He attended a press conference.' Although the meaning of these two sentences is very close, how can we use a program to score and judge?"
"We should compare the word vector corresponding to 'him' in the first sentence with each word in the second sentence to find the one with the closest meaning. And so on, 'media' corresponds to 'news'
Press conference' and 'publish' correspond to 'attend'. In this way, the similarity between word vectors is calculated and then merged into the similarity between the two sentences."
Lingzhi shows the similarity between the two sentences: 0.912. Everyone can easily understand how the similarity between the two sentences is calculated, because the distance between each word on the picture is very clear.
Then Lingzhi began to explain some technical details, including how to calculate the word vector of each word, etc.
As a lecturer, Ling Zhi knows very well that explanations need to be detailed and appropriate, so he does not elaborate too much on details and only uses diagrams to explain the core ideas of the article. Therefore, it is not easy to explain the ppt well. The details cannot be too in-depth, but it cannot be too general.
However, it is very important to grasp the degree.
After Ling Zhi finished sharing and said thank you, everyone in the conference room burst into applause unconsciously.
Ling Zhi sometimes noticed that when he spoke to the public, adding "thank you" at the end would make the audience applaud involuntarily. However, when everyone applauded him, they did not simply praise him, but they really felt that Ling Zhi spoke well.
"Yes, the experimental process and details are explained very clearly, the problems are well analyzed, and the paper is worth learning from. Okay, next, who else wants to talk?"
Ling Zhi breathed a long sigh of relief and sat down in another seat. A junior fellow student who had just been admitted to graduate school and came into the laboratory early stood up and opened his ppt.
"Um, hello, fellow senior brothers and sisters, today I want to talk about a paper, the title is..."
It was obvious that he was a little nervous, but Ling Zhi didn't care at all. He had come here in the same way.
"The algorithm of this paper is like this..."
"Wait a minute, I've told you before. Don't paste the English pseudocode of the algorithm listed in the original paper directly into the ppt. Who will read it if you do this? It's so complicated. You should draw a picture like Lexus.
, so that others will find it simple and easy to understand, you know? Don’t make me emphasize it again and again.”
"Oh, sorry teacher, I will pay attention to it in the future."
"Okay, then you continue."
So the junior brother finished his thesis with great trepidation, not knowing whether everyone understood or not. He was most afraid of the sudden silence.
Ling Zhi didn't think his junior brother was that bad. If he was bad, he wouldn't have sat there. However, it was just the first time he explained the ppt. He might not have considered the audience's feelings. Although he didn't understand much, he didn't really care.
Anyway, it’s not your research direction. If it’s really closely related to you, then reread the paper in private and understand it yourself.
I remember myself when I first entered the laboratory. At that time, I attended group meetings and felt dizzy listening to my brothers and sisters explaining ppts. After the group meetings, I frantically searched for information to make up for my concepts.
Thinking about it now, it's not that I looked down on myself at the time. In fact, it's important to understand the concepts that I didn't understand. It's just that my mentality at the time was too anxious, so I could have taken my time.
Unknowingly, the group meeting had already reached 11 o'clock. Lao Wang said a few more words, and the morning group meeting came to an end.