Returning to the camp from Lengshuixi, Zhou Zhi was preparing to make lunch.
Because there are almost no outsiders here, everyone is very at ease with the safety of the camp, and everything is thrown here, even three pack horses.
As a result, Zhou Zhi had just washed the vegetables and was about to cut the meat, he heard the wheat seedlings shouting in surprise. "What's wrong?" Zhou Zhi hurried over.
"The program is finished!" Mai Xiaomiao operates the monitoring screen on the computer: "Now is pulling the table to verify the correctness of the system operation."
"Eat meat and bran, just look at this shop!" Li Laosan rarely said rough.
However, after the results came out, the three of them were a little disappointed. The system analyzed the buying and selling points based on the stock price changes of the Dow Jones Stock Index from 1984 to 11 years, with a total of one million The high-frequency trading is simulated based on US dollars to verify the accuracy.
The system initially designed models in more than 50 categories, and was reduced to twenty-three by Mai Xiaosiaoyou. The main reason is that these twenty-three models have quick response, simple logic, simple learning, and accurate feedback. The other ones that are not easy to control through neuronal-like networks have been ruthlessly eliminated by Mai Xiaomiao.
Now the calculation results are out, and the final success rate is only 70%.
The three of them looked at each other. Although this accuracy rate has turned one million US dollars into 100 million US dollars in ten years, it is still completely unacceptable because those failed transactions will be ruthlessly Exposing the existence of this system in this world, thus arousing the vigilance of all forces.
"No..." Mai Xiaomiao muttered: "Why can't even win the simulation algorithm?"
The algorithm model of Baines Laboratory is a traditional data analysis model, and the success rate of the results is Ninety percent. The Mai Xiaomiao system was as high as 95% during static data verification, which is much higher than the data in Baynes Laboratory. Unexpectedly, after the addition of the time parameter became dynamic verification, it dropped sharply.
"Is it because of insufficient deployment of computing resources?" Li Laosan first thought of this question: "How about we expand the capacity again and run it again?"
" It's not that reason." Mai Xiaomiao shook her head: "After the elbow increased the computing power for Xiaozhi for the third time, we also used the gap time for the Shipbuilding Research Institute. Now Xiaozhi has become a computing power dynamic distribution system. , the utilization rate is much higher, which has exceeded 30% of the expected guarantee. "
Zhou Zhi's mind flashed: "Xiaomiao, count the time separately!"
"What does it mean?"
"We divide the time into three sections and count the system's transaction success rate in each year!"
As Zhou Zhi's words, Mai Xiaomiao's fingers A shadow flashed across the head on the keyboard like dancing, and followed a chart on the screen and jumped out.
"Yeah!" Mai Xiaomiao jumped up with joy when she saw the chart, hugged Zhou Zhi and shouted and jumped: "It's successful!"
"What does it mean to succeed?" Li The third brother looked at the crazy two people and felt that he was not involved.
Zhou Zhi held Mai Xiaomiao with one hand and pointed at the computer with the other: "Our system is a learning model. The theoretically, the longer it takes to process data, the more data it processes, the higher the accuracy of processing will be!"
"This curve is a time-sharing diagram of the hit rate from 1984 to 1995. The hit rate in 1984 was only 50%, and in 1989, the hit rate increased to 9%. Ten, surpassing the Baynes model's hit rate. By 1994, the data hit rate reached 95%. "Let's run again! "Mai Xiaomiao said: "Use the model you have learned now and then revisit it! ”
“Run and run! "Li Laosan immediately smiled happily and was full of expectations for the system: "I was shocked and I said how could Xiao Miao's system have such a problem! "
This is a bit brainless. Mai Xiaomiao quickly re-adapted the system and clicked to enter: "Okay, we'll start traversing the second time now. ”
This system was not deployed in Mai Xiaomiao's computer, but was deployed in the computing power center of Four-Leaf Crull Group. Here, Mai Xiaomiao only operates the system far away in Shudu through the satellite network and Fterm in the computer. "Are you still having fun?" Li Laosan looked at the two of them with a gaze. Now his mind was just about returning to the grain station as soon as possible. "Then let's close the stall." If you have reached this point, who still has the heart to continue playing, that's really heartless.
It was already evening when the three of them returned to the grain station. Mai Xiaomiao and Li Laosan got off immediately and ran into the big study room, reviewing the data results of the iterated new model.
"How is it?" After a while, Zhou Zhi also came in and asked the two of them.
Mai Xiaomiao retrieved a statistical sample of the new data. As expected, the data for the first three years have been adjusted, and the accuracy rate of reaching more than 95%. "This will be put into production!" Li Laosan was ready to rub his hands.
"You can take it easy." Zhou Zhi shook his head: "This data is not qualified yet, please continue to practice."
No mistakes, one song, one content, one in one 6 One 9, one book, one bar, one read!
"Bains Technology said that a 90% accuracy rate can be put into production." Li Laosan said, "Why are we still 95 now?"
< br>
"This has to mention the advantages and disadvantages of neuron-like systems." Zhou Zhi explained: "The strength of this system is the function of self-learning. As long as we design both the model and feedback algorithm, it can go more and more. The smarter the more correct the judgment. "
"But the flaw is that it has the probability of being exhausted like a human being. When encountering some situations that are beyond the norm or have not happened during training, it is based on The results calculated by our own 'experience' may be very different from the real situation we want. "
"In other words, this system may be judged 100 million times, but it does not exist to make us lose money. The probability of leaving 100 million times once."
"What should I do?" Li Laosan was stunned.
"Stern brother, don't listen to your elbows to scare you." Mai Xiaomiao was soft-hearted and couldn't stand Li Lao San's anxious. He quickly said, "The solution is very simple. Add the calculation mechanism, use our model to find the opportunity, and then use Bei to use Ens' model verification is correct and will trigger automatic transactions after passing."
"Then isn't this doubled the judgment process? It's just half the trading opportunities?" Li Laosan was a little unhappy.
"San brother, these are millisecond transactions." Zhou Zhixiao said: "In 38 milliseconds, we can theoretically trigger thousands of transactions in the global market. So what if we have less, what if we have to do with less. ? It's not that bad."
"Besides, we have to smooth the trading curve and not affect the convergence actions made by the stock market because of quantitative trading. Where is the timing of this loss? What you need is safety." (This chapter ends)