The blog has been inactive for a long time.
Although it’s not what I wanted, focusing on producing papers has made the world seem dull and boring to me. What water can flow out of a well that has run dry?
Fortunately, after a long period of wandering, I finally submitted this paper. Whether it is accepted or not, I can finally take a short breath. I also take this opportunity to reminisce about the past year.
Start my own research
In the second year of my PhD, I can finally start research that completely belongs to me.
For the first time, I have the freedom to completely control my own research. I am excited and confused. Excited because I finally have the opportunity to choose to work on problems I like in the way I like. Confused because I am not sure which problems I truly love. With this question in mind, I searched online and watched some relevant videos. Hamming said in the following video:
“Good research is the right time, the right problem, and the right way.”
However, to me, what is considered “right” is elusive. The definition of “right” is constantly being challenged and renewed. Academia iterates between being overturned and overturning itself. Newcomers bring new perspectives, and those perspectives reexamine old assumptions. Only work that has been repeatedly tested and changed an entire field becomes history and is written into textbooks. The big waves wash away the sand, and we are probably the ones creating the sand.
Therefore, I do not seek what is correct or the “right” way. I only seek “my way”. I hope that the methods I develop will be something that I personally like. If these methods can bear my personal mark, that would truly be something I cannot ask for.
Finally, I locked onto a problem and excitedly told my advisor. He said that the problem was very difficult. At the time, I didn’t take his words seriously. (Looking back now, he was right 😓.) I resolutely began to work on this problem.
Running into a wall
After identifying the problem, I divided it into several smaller problems and set several milestones. I thought each problem had a relatively simple and obvious solution. However, when I started to work on it, I realized that reality is always more complicated than theory.
The key to the problem lies in the subtle but significant gap between my abstraction and the actual behavior of the video codec and computer vision algorithms. To try to reduce this gap, I made countless attempts in optimizing the experiment, constructing the optimization objective, generating images, and optimizing methods. Every time before the experiment, I held hope, but after the experiment, I was forced to start over with disappointment.
I was caught in a delicate gap. If I found evidence that fundamentally denied the feasibility of this method, I would not have to be entangled with this problem anymore. If I eventually got readable results, I could move on to the next step.
However, obtaining fundamental evidence in computer systems research is extremely difficult. One counter-intuitive fact in computer science is that if experimental results do not meet expectations, the first thing to be questioned should be the experimental results. Most unexpected results are due to bugs in the program, and/or the abstract of the problem has omitted some critical parts of the system, rather than a fundamental issue that cannot be resolved.
It’s like walking through a maze. When walking through a maze, whether the maze has an exit is secondary. The most important thing is whether we still believe that the maze has an exit. When hitting a wall at each intersection in the maze, do we grit our teeth, reopen our wounded selves, and head towards the next unknown intersection, or do we look at our scarred selves and give up?
The only consolation on this road is that every step we take reveals something that no one else has seen before.
(By the way, every researcher must have a little bit of stubbornness and narcissism. Those without these traits cannot withstand the repeated beatings of setbacks.)
Giving up?
In the following months, I made no fundamental progress on this research. Time slowly ate away at my and my advisor’s patience. Finally, in early December 2020, my advisor and I had a conversation. He suggested that I consider switching back to the previous project and extending it further.
I also knew that extending the previous paper and building on its foundation would be relatively easier and could maximize the efficiency of going through the cycle of submitting, getting rejected, and resubmitting. From a standpoint of maximizing submission efficiency, I should indeed choose to return to the original problem.
And I should, if I view producing papers as my objective function to optimize my choices.
However, if I were to treat life as an optimization problem, it feels like something is missing. Like a cold pizza, it can be eaten but cannot provide warmth and hope to someone sitting alone in front of a computer in the middle of the night with a heavy heart.
Later, I realized that for me, the reason to continue with research cannot be based solely on efficiency or satisfying others, but also requires a bit of “I don’t care, I will do research anyway,” “I believe in myself,” a bit of obsession that just feels good for no reason, and a bit of beautiful, vivid, even mysterious faith, idealism, and heroism.
Yes, while the help of the advisor and collaboration with others are indeed critical on the research path, the most crucial thing is to embrace dreams and hopes. Just as in “The First Sequence,” the protagonist struggles for survival in the wasteland and at the end exclaims, “When disaster strikes, hope is humanity’s first sequence weapon.” As in “Spirited Away,” only by remembering one’s own name can one see through the illusion of magic and find the way back.
It’s okay to stumble step by step; there is a shimmering light ahead.
So, I told my advisor that I still wanted to give it a try. It was on that day that I started to find the missing piece of the puzzle, successfully performed the initial experimental results, and began the journey of submitting.
Writing a Paper
With the help of my advisor, we completed the experiment and submitted it to the Spring NSDI conference. Although we received good scores from the reviewers, it was ultimately rejected at the PC meeting. Since the experiment was already basically completed, I discussed with my advisor and felt that I could (and should) gradually start writing.
Through continuous writing and feedback, I began to understand academic writing better: academic writing requires constant questioning of oneself. What I am familiar with may not be noticed by others, and my personal opinions may not be convincing. It is necessary to clearly delineate the argument and present the reasons. After that, be cautious and repeatedly revise, fearing that an unfamiliar concept or vague argument will arouse the reader’s questions and confusion.
But the paper is 12 pages, 12 pages long. Every time I read what I wrote, I can always find something that others may not understand, and my advisor can always find a bunch of things that others may have difficulty understanding in what I thought was clear. It’s like the story of Jingwei trying to fill the sea. I work hard to carve out one sentence after another, immersing them in the vastness of 12 pages.
Writing has become a race against time, and I have included time in each small stone. When the deadline arrives, I have to stop, leaving all the imperfections behind.
Thick mountains could not stop the river from flowing into the sea.
If writing is just a race, even if it is painful, one can grit their teeth and finish the course.
However, in academic writing, in order to write good material, the writer must find objective, clear, and even constructive evidence for their beliefs.
This is like asking a game making team why they made this game, but the reasons cannot include love and liking, because those are not objective; asking a teacher why they work hard to teach, but the reasons cannot include the sense of achievement from seeing their students succeed, because that cannot be quantified. The reasons must be something like working hard to teach will increase the admission rate of students to top universities by xxx percent. It is also like asking an e-sports player why they train so hard, but the reasons cannot include that their goal has always been to win a championship (RNG loses, sad), because the championship is just a possibility, an upper limit, a vague promise. The answer must be something like without hard training, the success rate to perform a challenging but important game operation (like Vi’s Q-flash) is really low.
Academic writing does not welcome beliefs or vague reasons.
I cannot believe. I must search high and low to find the rationales.
I present each belief in front of me, and rack my brains to find ways to support them.
These beliefs were once the warm light that comforted me. But it seems that once evidence is found for a belief, it sheds its mysterious cloak and the light goes out.
So I disenchanted each small belief that supported me in facing one failure after another. Each belief was like a departing soul, going back to the moonlit mountains of Huainan, and returning to nothingness, unnoticed.
I remember my advisor once asked me when he was editing my article, “Do you subconsciously feel that writing is not worth spending several hours just to write a paragraph?” For me, this is not about whether it is worth it or not. Of course, I know that good writing is an important component of the road to academic success. But subconsciously, I don’t want to turn my beliefs into a cold, emotionless arguments. I don’t know if this change is good or bad, but once it happens, there is no turning back.
I also remember my advisor saying that when you feel that what you’ve done is flawed and imperfect, it is actually when you think through the problem and start to be convincing. That’s not far from publishing articles. However, how can something imperfect give you belief and warmth?
Unfortunately, I’m a PhD student. If I don’t publish my thesis, I won’t be able to graduate/find a good job, let alone becoming a professor. What can I do? I just have to grit my teeth and do it.
Outro
I started writing this short article in February 2021 with the original intention of writing a quick summary, but I dragged it out until now (November 2021) to have a relatively complete version. It’s just that when writing this kind of recollection, I need a lot of time to find the most important things to say or some consistent emotions, to avoid writing like a list. I’ve been busy doing experiments, writing, interning, and collaborating with others, and have more ideas than energy.
Unlike academic writing, there’s no external force driving me to write a blog. Therefore, each time I write, there’s a passiveness there. It’s not that I want to write, it’s that life makes me write. In other words, I need the primitive creative impulse and desire to express myself that life gives me. As for how to get this impulse, it’s similar to the principle of grinding away at a rough stone to make it smooth. It requires a bit of perseverance and adversity to help me organize scattered thoughts and emotions into a coherent thread, so that they can be formless but not scattered.
So, let’s leave the past year of me here. I’ve become a new me, and the old me has turned into words, which is good.