2024년 what I learned from the recent submission (CS PhD student)

 

 

몇일 전에 하던 연구를 풀페이퍼로 써서 서브미션을 했다.

 

이번 서브미션을 통해 배운 점 및 후기들을 기록으로 남긴다. 쓰면서 나도 다시 한번 되새기고, 다른 사람들에게도 도움이 될 수 있음 좋겠다. 같은 경험을 한 phd students에게는 웃픈 위안의 글이 될 수도  있겠다.

What I learned

Choosing a research project/topic is extremely important.

Choosing a research project/topic is extremely important. Probably it is one of the biggest factors that decides the end of the project. Once the project is on board and gains momentum, it is hard to change the track. Pivoting is not just making invested time succunt cost but also it discourages you and your teammates. If you need to pivot, do it when you explore research projects.

 

The second important decision point after the topic is the 'design' of the system.

The second important decision point after the topic is the 'design' of the system. The design means system design. Once you finalize the design of your system, the remaining works are technically materializing your design. What you are going to say in your paper and what you are going to do for the rest of the project will be justified. Oftentimes, you will find your self not being reminded of why you are doing a work. You start to have narrow view. At the worst case, you will find yourself spending too much time on not very important things. All of them can be answered by your 'design'. If a certain task is not really for your design, well then you have to step back and think about it again. Another reason that the design goal is so critical is that changing the design choice in the middle of the project is extremely costly. Basically, a lot of works that you've been doing so far could be useless and denied at the extreme case. You were also constructing the mental model and story based on the design. Changing the design means that you need to tear down large part of them and rebuild from scratch. Don't be cheap at investing your time on the design choice process. It will pay off eventually.

 

IS IT GOING TO BE IN THE PAPER?

You need to constantly ask yourself if what you are doing is necessary. How are you doing to tell something is necessary? It is as simple as it is. "IS IT GOING TO BE IN THE PAPER?" If no, stop doing it. If yes, you need to prioritze. To prioritize, ask "IS IT IMPORTANT TO PROVE THE MAIN THEME OF THE PAPER?" 

You have to continuously ask yourself this question. Probably every day and every week.

 

Give specific tasks and deadline when you allocate tasks your teammates.

You need to be smart at 'utilizing' your teammate. Your teammates' attitude/mindset will be "hey, what should we do?" They are waiting for you to decide and let them know what they have to execute. Unfortunately, in many cases, it is wrong to think that your teammate will be like "we are working together!". It is closer to "Hey, project leader, what should we do and what do you want me to do specifically?" and you must be like

"Let's do THIS, THIS and THIS. Priority will be THIS, THIS and THIS. Hey, daniel since you are good at this, you do this by this week. Hey, prajwal be cause you have done this before, you do this by next week."

 

and persuade them.

Another common mistake is you don't think of it as a situation where you need to convince them. 'They are teammate and why should I persuade them..' it is your project and to make your project be done, you need to 'utilize them'. To motivate them and let them work by the deadline, you have to persuade them with reasonable justification.

 

A key to this is being as concrete as possible. To get your paper out as quickly as possible, you must distribute the works. Teammates will not come and say 'I will do this task.' It will be the opposite. They are not going to do anything until you ask them to do a 'specific' work.

When the task is vague, what will happen is

1. you don't know what should be done for the project

2. However, you need some help and you feel like you are stuck

3. You give very vague instruction to your teammate.

4. ** the quality of your teammate's work will be very low. The time it takes for the teammate to do the task will be longer.

 

It is okay to make assumptions and simplify. You cannot solve all the problems.

In research projects, I need to simplify the stuff. It is okay to make some assumptions. I cannot solve all the problems. That will be insanely risk attitude. You are a PhD student. You will do all the work and there is a certain limit that one person can do. Make assumption. Your paper will not solve all the problems.

 

Try to do the project the solution would be simple. The problem is okay to be complex and niche. But the potential solution and technique that seems to be required should be simple. Otherwise, you will spend a lot of time on engineering things. I don't think it is ideal unless that engineering skill is directly useful to get a job. Just doing software engineering should be avoided since you will not have comparative advantage against undergrad.

 

Find as niche problem as possible for your project.

Find as niche problem as possible for your project. I feel like they appreciate you if you solved niche problem. Project that is trying to solve general problem is difficult to be appreciated. (do application-specific system project......)

It is also easier to bring interesting insight people will be curious of.

 

Overall reaction

Overall, I am happy with my submission. We could have started shortening earlier. (we shortened 1 page(double column) at the last 30 minutes). However, still content-wise, I am generally happy. I(We) did what we were supposed to do. I worked hard and for the last one month, we made decent progress by focusing on what we ACTUALLY need.

 

 

Plan after this

2년동안의 연구를 어느정도 한번 매듭을 지은 상황이다. 지금 하고 있는 연구 주제에 대해 믿음이 크지 않다. 피벗을 고민하고 있다. 중요한 트렌지션 포인트, 디시전 메이킹 포인트이다. 이번이 마지막 피벗을 할 수 있는 기회이다. 결정을 잘내려야한다. 지친 몸과 마음을 정비하고 잘쉬고 운동도 하고 다음 연구 프로젝트를 잘 고민해봐야겠다.

 

새로 알아보고 싶은 것은 따로 글을 만들 계획이다.

 

논문 쓰면서 먹은 치폴레들... almost everyday. One chipotle for two meals to be efficient....

(and I lost 4 pounds. not surprising)