Finally, I can take a breather.
Friends who follow Scientific Space may have noticed that updates have been quite slow recently. This all dates back to this year’s winter break...
At the end of January this year, due to various reasons and combined with my own interests, I looked for an internship. The position was for Python programming. The job was posted on the forum of South China University of Technology. The description was quite concise, and I submitted my resume just as concisely. To my surprise, I received a reply and was hired. I started working in February. After joining, I realized that the company is actually a well-known domestic e-commerce enterprise, and my main task was data mining. Although I had some experience with Python, I was basically a novice in data mining. Therefore, I had to learn while working, frantically catching up on data mining knowledge. During this process, I learned a lot. Before this, I didn’t even know what a “feature” was, nor did I know about “Logistic Regression” or “SVM”... I was completely ignorant back then.
Soon the new semester began. Since there are fewer courses this semester, I have been able to maintain a state of attending classes while interning, which continues to this day. After school started, I wondered if the data mining knowledge I learned at work could play a role in my coursework. Consequently, I participated in several competitions and projects related to data mining, including: the “Liangjian Cup” of the School of Physics and Telecommunication Engineering, a scientific research project from the School of Mathematical Sciences (Employment Demand Analysis Based on Data Mining), survey data analysis for the School of Information Technology in Education, and the Teddy Cup National Data Mining Contest (which I also participated in last year, winning only a consolation prize). These four projects, combined with work and my regular curriculum, came one after another, making this semester exceptionally fulfilling. I have also benefited greatly from them. Today is the submission deadline for the Teddy Cup paper. When I uploaded the paper and clicked “Submit,” the thought “Finally, I can take a breather” surfaced in my mind.
Of course, hard work pays off, and I feel I have gained a lot throughout this process. On one hand, through this journey (work + competitions + projects), my ability to process data with Python has improved significantly, and my understanding of Python has deepened. On the other hand, what impressed me deeply was that in two of the data mining competitions, I learned about an algorithm called “Deep Learning” (related to multi-layer neural networks) and cutting-edge data mining techniques like “Autoencoders.” In fact, when I was learning algorithms like Logistic Regression and SVM, I was very confused: why are there so many algorithms? Why isn’t there a universal and effective algorithm? Why do these algorithms work? I wasn’t very clear about these questions at first. It wasn’t until I encountered neural networks and deep learning that I found a satisfying answer. Neural networks (deep learning is a generalized neural network) are exactly the kind of universal and effective algorithm I was looking for. Its principle is to use multiple composite functions for fitting. If I have the chance, I will share my experience with neural networks in another article.
Here, I am willing to spend a bit more space talking about deep learning because it is truly worth understanding, even if only at a conceptual level. It is said to be the algorithm closest to Artificial Intelligence (without exception). The reason it is called universal is simple: it originally simulates the human thinking process (problems are posed by humans, and humans can think; if an algorithm can simulate humans, then the algorithm itself must be very powerful). A very important characteristic of human thinking is abstraction. You might not have noticed, but the process of “abstraction” is actually a process of information loss—it’s just that the lost information is what we consider unimportant. For example, if we study basketball, volleyball, and football, we can abstract some of their overall information, such as they are all spheres, they are all sports equipment, etc. Such an abstraction process allows us to recognize the commonalities of things while reducing the brain’s workload and improving processing efficiency. For computers, deep learning algorithms implement exactly this process! Through multi-layer neural networks, they construct autoencoders to simulate the human “abstraction” process! (Of course, in such a short space, the reader only needs to understand this concept.)
Half of this semester has passed, and the bulk of the data mining content is also behind me. So, for me, what is the next step?
It is time to return to my mathematics and physics!
Although deep learning is interesting and promising, the people I admire most are still those who proposed the deep learning algorithms. When learning deep learning, I also tried my best to understand its essence from a mathematical perspective. Therefore, I love mathematics more; I am more willing to engage in theory and continue with the mathematics and physics I love. In the process of thinking about mathematical and physical theories, I can feel an indescribable sense of pleasure and achievement. I prefer using a pen and a piece of paper to construct my ideas, rather than having to face a computer to complete them (although doing math and physics inevitably involves computers, that is a different feeling). I am not sure what kind of job I will choose in the future; it is even possible that I will work in data mining or machine learning. But no matter what, that love for mathematics and physics has always been there.
Therefore, it is time, to continue my study on science.
When reposting, please include the original link: https://kexue.fm/archives/3319
For more detailed reposting matters, please refer to: Scientific Space FAQ