Rebirth: I carry a desktop computer inside me

Xu Wei did not give them a chance to fight, and a faint sentence had already sealed Zhang Jing's thoughts.

The strict requirements of the big guy, he knows, and he is not less scolded for this.

But Jiang Cheng can get this evaluation from Xu Wei, which is enough to prove that Jiang Cheng is much better than him.

Although Jiang Cheng is a direct doctoral student, one class later than him, it is possible that Jiang Cheng graduated with a doctorate, and he is still studying hard for a Ph.D.!

Seeing that everyone was no longer arguing about this matter, Xu Wei directly began to arrange tasks.

“...... We know that biological neurons, especially dendrites in the human brain, have extremely complex topologies and dynamic processes that allow the brain to process complex tasks while maintaining low power consumption..."

On the small whiteboard, Xu Wei began to explain the project.

"Then, our task is to find ways to develop such a new type of memristor to store information when the current passes, and at the same time have computing functions to achieve the integration of storage and computing."

In this regard, Xu Wei studied earlier and deeper, so everyone listened carefully.

"Before that, I had done some research on memristors, and I found that to achieve the memristor element we need, we must first overcome the influence of bit error." Xu Wei continued, "Then, we can borrow the weight and residual calculation in the neural network, and this piece, Jiang Cheng's level can be compared to Lao Luo, led by you."

"Okay." Jiang Cheng nodded, did not hesitate, and directly answered.

Through Professor Xu Wei's arrangement, the five people were initially divided into labor, of course, there will be many cooperation parts later.

This requires Xu Wei to specifically coordinate and participate in each step to discuss how to solve the problem of R&D.

This project, big. Compared with Jiang Cheng's previous research, it may only be a small part of overcoming the bit error.

But on Xu Greatman's side, it is the entire chain of plans realized by the entire memristator.

Including algorithm implementation, memristor chip development, circuit design and so on.

Any section of the chain that goes wrong will affect the progress of the overall scientific research project.

"Okay, do you have any questions?" Xu Wei erased the small whiteboard full for the 31st time.

However, everyone took notes and wrote down what they were going to do.

Jiang Cheng raised his hand.

Xu Wei frowned slightly, "Xiaojiang, what's wrong with you?"

"Mentor, is the project short of scientific research funds?" Can I fund in the name of a company, that is, to require the sharing of results and patents. Jiang Cheng asked.

The corner of Xu Wei's mouth twitched.

"No, but how much can you fund?" Xu Wei gave him a blank look.

"20 million in the early stage, not enough later." Jiang Cheng smiled slightly.

Depend on! Evil capitalist.

Even the other PhDs were a little side-eyed.

I heard before that the entertainment industry has also brought funds into the group, why have our scientific research projects also become popular?

Is such a rich big moneyman?

To be a team deputy, that's a must!

Xu Wei was silent, "I will discuss this matter with the leaders of the department, but there should be no problem."

Jiang Cheng nodded, a little happy.

What else to discuss, your Director Xu is also the leader of the department, the deputy department head!

----- (⑅˃◡˂⑅) ----

sure enough, Xu Wei said the next day that the department agreed.

Jiang Cheng himself is also engaged in research and development inside, really produced results, and also has his own strength, not to mention that he is willing to pay so much money to fund, which can save a lot of funds in the hospital.

He gave too much.

Jiang Cheng was also very cheerful, and after signing the agreement, he arranged for the group's financial staff to call the Institute of Microelectronics 20 million yuan on the same day to specifically fund the memristor project.

Suddenly, Professor Xu Wei, who had a lot of money in his hands, took a big step.

If it is school or financial support, it is usually a few million at most.

How can Jiang Cheng be so generous?

So, what equipment is needed, buy!

What materials are needed, buy!

Summer vacation, it's hard to find miscellaneous players, it's okay, let's spend money, give more subsidies....

That is, Li Xinyi, Tang Jianshi and Song Lin communicated with Jiang Cheng more politely, some things that need to be coordinated and coordinated, Jiang Cheng as a deputy, although he is only a straight doctor, but he feels very comfortable doing it.

There was no resistance.

The biggest resistance comes from the project research.

The theoretical design is beautiful, but in fact, in the process of research, many difficulties will be encountered.

The design of the memristor is not carried out according to the chain, such as Jiang Cheng applied the knowledge of neural networks to design convolutional pools and residual blocks, which are used to overcome bit errors, but in the development of memristor materials and circuit design, it is also necessary to keep pace.

Any problem on the chain is easy for the entire chain to need to be adjusted together.

For nearly twenty days, Jiang Cheng followed the team to focus on this memristor and continued to study it.

This made Luo Ruihua a little strange.

Jiang Cheng, who has always been diligent and studious, didn't see it at school.

Could it be that you went home during the summer vacation, or were you all busy with the company?

Are you thinking about that big issue?

Fortunately, it was the summer vacation, Luo Ruihua planned to rest in addition to a meeting, and there was no phone to disturb him.

In the past 20 days, with the help of Professor Xu Wei and Dr. Tang Jianshi, Jiang Cheng combined with the convolutional neural network algorithm to propose a 3P hybrid training model.

That is, it is composed of off-chip pressure training and on-chip adaptation, and the device error, circuit error and inter-array error are mixed and trained, and a suitable systematic error model is introduced by Jiang Cheng through hundreds of millions of simulations of the in vivo computer.

Coupled with weight mapping, multiple key layer weights are adaptive training to continuously improve accuracy and overcome bit errors.

In the vernacular, it is to design multiple nodes and mix to judge right and wrong.

If there is an error, even if a node passes, but most nodes will be wrong (because they have received professional training), then for the wrong information, discard and retain the correct information.

This 3P hybrid training model has been highly praised by Xu Dawei.

It is believed that even if there will be some changes in the subsequent design, this model does not need to change much, and continuous training can be done.

Accuracy will continue to improve.

At that time, when the memristor is stored and calculated, it will be implemented theoretically without any errors.

This is the key to securing information storage and preparing for computing.

In the first chain, Jiang Cheng has taken the lead in breaking through, which makes the team very excited, and decides to take a day off and tackle tough problems again.

Even if they read books and have weekends, they work hard as a team, and they really need to rest.

Constant tension of nerves can be exhausting.

On July 21, which also happens to be the day when Jiang Cheng is resting, Maple Fire's mobile phone EX1 will hold an offline press conference at 3 pm.

The address is Hall 1 of the China National Convention Center.

In the hall, 103 media and self-media companies and 100 fans were invited, as well as Li Xiao, president of Taomao, Hu Yihua, president of Huiding Technology, Li Shaoyang, manager of the payment department of Ascendas Weixin, and some partners, in addition to Yu Xinran, a subsidiary of Maple Fire Group.

Due to the relatively large scale, some business bigwigs were invited.

Even if Xu Wei didn't take a holiday, Jiang Cheng had to take a leave to come.

This is the first step for Jiangcheng Fenghuo mobile phone to step into the high-end, and it is also Jiangcheng's first self-developed SOC chip release conference.

As it should be, he was to be there.