Time Travel: 2014

Chapter 163 The top scorer in the National College Entrance Examination (7)

About future plans.

Lin Hui has actually made quite clear plans.

And it's not just talk on paper, Lin Hui has already put it into practice.

At least Lin Hui thought he was actually moving forward steadily.

Although it seems that Lin Hui's steps are not very big now.

But from a macro perspective, or from the perspective of others, Lin Hui's progress is actually not slow.

To sum up, Lin Hui has achieved some success in academic field.

More specifically, Lin Hui has made some achievements in the generative summary algorithm.

As for why we should make a fuss about the generative summary algorithm.

It was because Lin Hui regarded the generative summary algorithm as a stepping stone to enter the game.

(The reason why I entered the game this way is naturally because the threshold for the field of natural language processing to which generative summary algorithms belong is relatively low.

By analogy, natural language processing is to neural network learning much like number theory is to mathematics.

You can't say that number theory in mathematics is unimportant.

But the threshold is really low.

Of course, although the threshold for number theory is low, the upper limit is very high.

The science of natural language processing is also similar.

The threshold for natural language processing is also very low.

When it comes to language, everyone knows language, and everyone has a certain ability to summarize and refine language.

After all, those years of Chinese language learning were not in vain.

It can be said that compared to some things that are confusing even for beginners, the threshold for the subject of natural language processing is also very low.

The threshold is low, which means that the access rules are not high. If the access rules are not so troublesome, Lin Hui will not be taken too seriously even if he enters the game.

Although natural language processing has a low threshold.

But like number theory, the upper limit is high.

What does a high ceiling mean?

A high ceiling means that you can reach some very high realms.

And being able to reach a very high level means that you can easily break through in the future! )

And how does Lin Hui plan to break the situation?

Generative summarization algorithms are essentially the product of the development of deep learning.

And now at this time point.

But only neural network recognition.

There is no real sense of deep learning.

Lin Hui knows very well what is the key to unlocking real deep learning?

——Residual neural network.

(The concept of residual neural network was proposed by a Microsoft R\u0026D team in 2015.

The residual neural network is a neural network that relies on the residual learning framework to ease network training.

It is different from previous neural network architectures.

The architecture of residual neural networks turns the layer into learning a residual function with respect to the layer input, rather than learning an unreferenced function.

Empirical evidence demonstrates that these residual networks are easy to optimize and can significantly increase depth to improve accuracy.

Lin Hui remembered that a previous research team evaluated a residual network with a depth of up to 152 layers on the ImageNet data set.

What is the concept of this 152-layer depth?

In the past, when we were crazy about scrolling, it was common to use servers to run neural network learning models.

But those days of crazy rolls are history.

As far as the current time and space is concerned, the depth of 152 layers can be said to be beyond the depth that can be achieved by various current neural network learning models.

In terms of quantitative analysis, the depth of 152 layers means that it is 8 times deeper than the depth achieved by the current mainstream neural network learning in this space and time.

Greater depth often means greater accuracy.

Now in this time and space, it is also clear that deeper depth is more beneficial to the engineering surface.

However, blindly increasing the complexity of traditional neural networks has led to shortcomings such as a sharp increase in model complexity.

And this highlights the value of residual neural networks even more.

Because in terms of complexity, residual networks tend to have lower complexity than traditional neural learning networks.

Furthermore, because deep networks can more naturally integrate low/medium/high-level features and classifiers in an end-to-end multi-layer manner.

The "levels" of features can be enriched by the number of stacked layers.

It can be said that the residual neural network has overwhelming advantages compared to the traditional neural network model.

It is precisely because of these and other advantages that the residual neural network in the previous life was quickly and widely used as soon as it came out. )

Residual neural network may be an obscure term to outsiders.

But in fact this is the key to the future.

And it is in Lin Hui's hands.

Lin Hui's metaphor is not excessive at all.

Only after the emergence of residual neural networks did the concept of deep learning really appear.

Specifically to the application level.

After the emergence of residual neural networks.

It can lead to a series of breakthroughs in the field of image recognition.

And a breakthrough in the field of images.

It is likely to lead to the development of [face recognition technology].

What's more, Lin Hui still has some information about the development of facial recognition technology from his past life.

This undoubtedly means a great advantage for Lin Hui.

The emergence of deep learning not only promotes the development of image recognition.

After the emergence of deep recognition technology.

The rapid development of [artificial intelligence] has truly laid the foundation stone.

To a certain extent, using generative summary algorithms to exchange money is no longer Lin Hui’s highest-level appeal.

Lin Hui now has bigger plans.

Whether it’s artificial intelligence or facial recognition.

Both are nearly a trillion-level market!

However, a journey of a thousand miles begins with a single step, and you still have to work hard.

If the foundation is not stable, you will easily fall when you climb high.

When Lin Hui was thinking about the trillion-dollar market.

Members of the discussion group are still discussing future professional choices.

He Siyuan: "I haven't thought about it yet...my mother suggested that I study accounting."

Liu Yao: "Listen to Auntie, you won't suffer any disadvantages when learning accounting..."

He Siyuan: "Well, I feel that accounting work is very boring."

Liu Yao: "But you can earn a lot by studying accounting. If Shuimu comes out, it will be easy to enter the four major universities in the future..."

He Siyuan: "[Doubtful][Doubtful]What are the four majors?"

Liu Yao: "Uh... go ferry yourself..."

He Siyuan: "I thought it was better than an organization?

How many law firms can be ruined after only half a day? ?

Are you kidding me? Not interested in!

I think no matter what you study, you still need to study science and technology. "

Liu Yao: "Come on, how much money can I make by learning technology?"

He Siyuan: “It’s vulgar, but it’s cool to learn skills!

Making money is a temporary thing.

Being handsome is a lifelong thing. "

Liu Yao: "It's over, Lao He is destined to leave us..."

He Siyuan: "What do you mean?"

Liu Yao: "Since you are engaged in technology, you will most likely have to go abroad in the end..."

He Siyuan: "No, after all I have received 15 years of patriotic education, it is impossible for me to go abroad."

Cao Haichao: “This has nothing to do with being patriotic or not.

I think as far as general technology is concerned, the technology is not the same wherever it is done? "

Tap the screen to use advanced tools Tip: You can use left and right keyboard keys to browse between chapters.

You'll Also Like