Reflection on My Last Semester and What I Wish for my future
October 2019
​
Looking back on what I've written half a year ago and it seems that many things have changed dramatically. First of all, I found the summer research at UC Berkeley where I spent a meaningful summer there. I was very lucky to be admitted to Professor Dawn Song's group and did some research in the security-related issues in Deep Learning. To be honest, before I found the chance to the summer research, I was worried that I might not even find one. However, I was honored to have the opportunity to learn from brilliant people and have some research experience in security-related topics. Though I didn't do research in this area previously, I think this area is very interesting and meaningful, since robustness is still of great concern in the deep learning systems. It is exciting to attack the algorithm using backdoor trojans or adversarial examples, while it is also challenging and therefore fulfilling to build really robust systems, which needs substantially more efforts and comprehensive thoughts. I learned a lot during the summer research from every person I met, especially Prof. Dawn Song and Prof. Bo Li, and I felt I had substantial progress and gained more insight into my research. The other thing is that I finished a project with Prof. Tong Yang on a generic mechanism which can be applied to counters. I gained much confidence through this project and I conquered a barrier which seemed to be hard for me. I know clearly now that I have the ability to do well in this field. Finally, I went back to have my internship, where I will continue working on the projects which I think are very meaningful.
Soon I will graduate from my school and now I'm applying for graduate school. No matter where I will go, I hope I can learn something every day and make some real contribution one day.
About My Research Experience
March 2019
​
In the common sense, my research experience wasn't very pleasant. When I was a sophomore, I followed Professor Yasha Wang to analyze medical data. I read many papers then, but I found that I had no idea about data mining at that time. Frankly speaking, I thought I lacked the ability of self-study at that time, so I found machine learning difficult at that time.
Anyway, I then followed Professor Tong Yang to do data stream processing using probabilistic data structures since September, 2018. This group was highly motivated and this group indeed produced many papers, and I also learned a lot during the projects, but this group was so large that there were not many chances for me since I was a beginner then. Our group's aim last year was two Sigcomm papers. Since I went to this group, I had worked on mathematical proof of one of the papers. I actually did work out the proof, but the algorithm changed afterwards. Two other students and I worked on the new mathematical proof again, but we didn't make it, and I wasn't able to contribute to this paper. What's more, I soon found out that I lost interest in this area, and I got really frustrated at that time, due to both the failure and some personal affairs.
I started to reflect on my past research experience. I was told that I lacked confidence, and I think that's true. Besides, I didn't make a wise decision in the last two research experience, because I didn't figure out what I really want either. I decided to make a change. In the winter break, I began to read papers about Bitcoin and Arvind's Bitcoin and Cryptocurrency Technologies. Though I think I won't do research about Bitcoin recently, I find the topic really interesting. At the beginning of this semester, I began to study deep learning again. I took two courses in Computer Vision this semester. I read Ian Goodfellow's deep learning, and I self studied deep learning by Andrew Ng and cs231n in three weeks. I took the interview at SenseTime recently and I am now an intern at SenseTime. At this time, I feel very confident about my future research. Maybe I am not well qualified at the time being, but I think I can catch up in the following days. I think this is the most fulfilling month since I went to college.
In fact, I wasn't looking for internship in the area of deep learning previously. I'm hesitant at the beginning, because I think I'm not qualified enough. However, now I think I can be helpful. I sincerely hope that I can get the chance to get everything on the right track.
​