I'm a scientist, why can't I have faith?

If mathematics is the underlying logic of artificial intelligence, then code is the language for communicating with artificial intelligence.

Only after laying a solid foundation in both can one truly step into the field.

Machine learning, deep learning, natural language processing, computer vision, digital image processing...

These are the most important aspects of AI; in comparison, the previous content was more for the convenience of teaching.

Even if Wang Dong is a genius, he cannot master all of the above content in this short National Day holiday, but he can still learn about it.

This extraordinary learning ability even impressed Mr. Turing.

Alan Mathison Turing praised your belief in artificial intelligence.

[Gain Focus*0.1]

After a holiday of diligent study, Wang Dong felt he was now qualified to open the shared document sent by Su Yuran.

The document is very comprehensive, containing not only information about the competition but also the preliminary algorithms to be submitted.

It seems that Su Yuran really trusts me, sending over the preliminary algorithm without even bothering to keep it confidential.

However, looking back, it all started with the email he sent to Academician Du Fan. Coupled with the resemblance between Su Yuran and Du Fan, Wang Dong suddenly felt that things were not so simple.

Senior... seems to have an unusual background.

That's true, they don't seem to need to keep it particularly secret, after all, it's just a small competition, not a paper to be published in a top journal.

With a sense of awe, Wang Dong began to browse through the various contents.

Although this is the first edition of the "East Asia Smart Cup", its competition system is already very well designed.

The preliminary round adopted a common solicitation strategy, where the organizers announced a real-world application scenario for AI, and participating teams only needed to submit their specific algorithm solutions to the organizing committee before the deadline to complete the competition.

Then, through discussion and screening by the organizing committee, the list of participants for the finals was determined.

In the final round, the problem is announced on the spot, and the team has to come up with their best solution within a limited time, and the score is determined accordingly.

The topic for the preliminary round was "Design of AI Algorithm for Emergency Dispatch of Sudden Urban Traffic Congestion," which is a very common use case.

The problem requires balancing congestion prediction accuracy, emergency dispatch latency, and system fault tolerance under the constraints of 30 days of historical checkpoint data and real-time accident flow.

Although Wang Dong has never participated in competitions, he has undergone intense exam training in his senior year of high school, so he has a lot of experience in solving problems.

It's clear to anyone with eyes to see that the accuracy of congestion prediction, the delay of emergency dispatch, and the system's fault tolerance rate are the three core indicators that are key to the preliminary evaluation.

In other words, the algorithm design should revolve around these three metrics, and the final performance of the algorithm based on these metrics will be the key to whether it can advance to the next round.

Sure enough, that's exactly how Su Yuran and her team designed the algorithm.

It took Wang Dong an entire afternoon to carefully study Su Yuran and her team's preliminary round answers from beginning to end.

Judging from the algorithm test results attached at the end, this is a highly complete algorithm.

Not only did it eliminate 99% of redundant variables, but operator fusion and model quantization reduced inference latency to 82ms, and even boosted prediction accuracy to 98.7%, with a fault tolerance rate of 0.17%.

For passing the preliminary rounds, it's more than enough; it could even be described as a game-changer.

However, in Wang Dong's eyes, who possesses both mathematical logic and the Turing Test, there are still areas for improvement.

The core architecture of this algorithm design lies in the globally weighted spatiotemporal cross attention and dual-path redundant inference branches. It is these two modules that bring the accuracy and fault tolerance of the entire algorithm to such a high level.

However, the designers overlooked a crucial point.

Global weighting can indeed improve overall accuracy, but sudden accidents are typical of long-tail distributions.

Using global weights dilutes the features of extreme events, even though three more layers of regularization were added later to compensate for this lack of precision.

However, the fit of edge scenarios was actually lowered, and overfitting and risks were all hidden in the extreme conditions not covered by the test set.

Dual-path redundancy was chosen to meet the fault tolerance requirement by running in parallel with dual paths throughout. However, in emergency scenarios, the fault tolerance of single-path inference is sufficient for 94% of normal operating conditions. Running in parallel with dual paths throughout would only needlessly increase latency.

Although these are all minor details, the modifications may elevate the algorithm to a higher level overall.

Even though the algorithm would pass the preliminary round without modification, Wang Dong still left his thoughts in the document.

This can be considered a team contribution.

In Wang Dong's social philosophy, only those who contribute to the team can better integrate into the team.

He didn't want to be seen as a parachuted-in relative and then ostracized.

"This way, I should be able to integrate into the team more quickly..."

With his high-speed train departure time approaching, Wang Dong wrote down a rough outline of his thoughts on the document's message board before rushing off to the train station, completely forgetting about signing his name.

By the time he arrived at platform 7 of Tianhai West Station, there were only two minutes left before departure.

Fortunately, it was the starting station; otherwise, whether I could have boarded the train would have been another matter entirely.

As the Fuxing bullet train slowly started moving towards the Yanjing Expressway, the National Day holiday officially came to an end.

The freshmen have now passed their "newbie protection period," and what awaits them now is the real test.

Wang Dong leaned back in the comfortable seat of the Fuxing bullet train, watching the lazy sunset outside the window, and unknowingly drifted into a warm dream.

Meanwhile, in a laboratory at the School of Information Science and Technology at Yenching University, Wei Linfeng had been so worried that he hadn't had a break for several days.

He spent the entire National Day holiday in this small room of less than 30 square meters.

For graduates, you have to give up your rest time to complete your graduation project, which is also a way to prepare for the grueling life of graduate school.

After all, graduate students don't have winter or summer vacations.

Wei Linfeng is a senior student in the School of Information Science and Technology. Although he has been admitted to the graduate program of Academician Du Fan through the recommendation system, it does not mean that his life will be easier from now on.

The graduation project at Peking University is notoriously demanding; failing to meet the requirements prevents normal graduation, even for graduate students who have been accepted into academicians.

He didn't want to change the course of his life because of this.

Therefore, Wei Linfeng began preparations well in advance.

He originally didn't plan to participate in this year's "East Asia Smart Cup," since winning the award wouldn't help him at all, except for the prize money.

But to my surprise, the topic of the preliminary round coincided with my graduation project, plus the invitation from my junior, Su Yuran.

Everyone at Zhengxin Academy knew that Academician Du Fan was her uncle, so Wei Linfeng naturally accepted the invitation from his future mentor's niece.

It's always better to have someone help you with your graduation project than to work on it all by yourself, isn't it?

Moreover, with an academician's endorsement, my graduation project might be able to be completed ahead of schedule.

Although everyone on the team was very capable, and Academician Du Fan also offered some appropriate suggestions on the algorithm, the results were more than enough to pass the preliminary round.

However, it still falls short of Yenching University's graduation project standards.

The rest of the journey is up to Wei Linfeng himself.

As a result, he spent the entire holiday in the lab searching for literature, but couldn't find any clues about optimization.

As if by some strange twist of fate, he opened a long-dormant shared document, which contained the entire design process of the algorithm from scratch.

Wei Linfeng hoped to find some inspiration from this.

Then he saw a message that had appeared out of nowhere.