Well, it’s my first time to write an article online. This is probably the first review I’ve written since graduating from junior high school. There are two reasons why I came here to write an article: First, I feel that I haven’t systematically summarized my thoughts since I was in college.
Combing and implementing it into articles has greatly increased the inertia of thinking and failed to cultivate the habit of systematic thinking, which is a pity. The second is the use of Markdown writing and typesetting method. I have only recently learned about the advantages of this writing method. I would like to borrow it.
This opportunity to learn.
So from now on, try to develop the habit of recording your thoughts here.
Closer to home, if you are a liberal arts student, I suggest you try reading this book when you have time. It is not just for engineers. This world is not only full of romance and chicken soup, but also some things worth pondering and exploring.
When I was in college, I developed a strong interest in stock speculation. Every day I thought about how to get rich overnight in this almost zero-sum game market. I checked a lot of unknown people’s various tips and tricks. As you can guess, the result was nothing.
Useless.
It wasn’t until I read several books about how American hedge funds operate that I suddenly realized: Damn it!
The most powerful stock market participants in the world no longer rely solely on the human brain’s intake and judgment of information to make investment decisions. They rely on advanced computers and trading models to make profits. If you are also interested in this field
If you are involved, you should have heard of the famous name "Renaissance Technology Company", and the core of this advanced trading method is: mathematics.
Or, more closely to the times, it’s called big data.
Mr. Wu Jun combined his profound mathematical skills and long-term project development experience in first-tier Internet companies such as Google and Tencent to use extremely simple language to describe the search, translation, navigation, speech recognition, web crawlers, web page rankings and processes we use in our daily lives.
The mathematical principles of anti-cheating and other Internet functions provide an approachable science popularization of work that in the eyes of ordinary people only belongs to engineers and scientists, often using only one mathematical equation to reveal many terms that we feel are high-end in daily life, such as "
Artificial neural network", "information entropy", "Bayesian network" and so on.
For me, I read it with the mentality of exploring interest at first, but I didn’t want to accidentally read some content related to automated trading.
Like the maximum entropy model, Markov chains, Bayesian networks and artificial neural networks have also made great contributions to today's automated trading. These mathematical models and ideas have gradually entered the U.S. investment market since the 1990s, and they have achieved
Even companies like Berkshire Hathaway are far behind.
The information that a human brain can process is ultimately too limited. Even the collective wisdom of several traditional fund managers cannot match the energy of the entire market.
However, with the help of mathematics and today's powerful data acquisition and computing capabilities, it is possible for us to quantify countless influencing factors and make accurate judgments.
Another question that makes me think about is how to train the parameters of a mathematical model without data or only a small amount of data, and use data to promote function iteration.
Because we encountered a problem at work recently, the development of a product in our hands is almost completed, but the core functions of this product require a batch of data. If we have enough user traffic, we can continue to iterate based on user data feedback.
The core functions of the product, however, due to the design issues of our product, if we completely rely on user-contributed content (UGC), it will seriously affect the user experience. This is a cold start problem.
We act as the first batch of users (or hire people) to do UGC in corresponding scenarios, and the cost is too high.
This forces me to think about whether it is possible to use certain algorithms to improve the product as quickly as possible with less data.
"The Beauty of Mathematics" also gives some cases of "making things out of nothing", such as the PageRank algorithm that helped Google become famous.
Finally, describe two simple concepts.
What is encoding and decoding?
During the time when I first started working, as a liberal arts student, I had a lot of headaches about these two concepts and related issues. In this book, I found the answer.
To give a popular example, when we express what we think in our minds in words, this is encoding. When a person listening to us absorbs what we say and understands it in his head, it is decoding.
Maybe it seems to you that this is a very common process, isn’t it a very natural thing?
If you think about it carefully, why the things you think about in your brain can be spoken through language or written down through words. The information stored in spoken words and words written on paper are completely different from the information stored in your mind. This is
A set of conversion rules is actually encoding and decoding, and English and Chinese are two different sets of encoding and decoding rules.
Similarly, when we make a phone call, the acoustic information we emit needs to be converted into electrical signals, transmitted to the other party through radio, and then converted into acoustic signals that humans can understand. This is also a process of encoding and decoding.
The encoding and decoding of all forms of information are essentially mathematical tasks.
A more direct manifestation is that when we type information through a computer, the most common way is to type. However, when this information is encoded and handed over to the computer, it is stored and transmitted in binary.