The first idea is: Perfect papers — they don’t even exist in the world.
But let’s try our best today.
The have the two portions: Writing perfect papers and Writing good papers fast!
Story
How do I make a great movie?
A great movie starts from a powerful and convincing story.
A great app starts from a tasteful design.
Three essential elements of a great story.
My solution, My math proofs, My simulation results?
Not really.
This it try to show is not like I am very smart, I’m smarter than you think, or the thing actually works.
Your paper is not really the vest place to show your theoretical prowess.
It’s about advancing the state-of-the-art.
The problem (what are we solving?), What’s new (what’s really new in your contribution as compared to related work?), How is it better? (how is your contribution better than the state-of-the-art in the related work?)
The problem
Doesn’t have to be something “trendy”.
That’s a good idea, that’s a good starting point, but it doesn’t have to be something trendy.
But it has to be something you really enjoy working on.
The “trend” may not last forever.
P2P model, something it goes down, but something iit goes up in the 20 years.
Timing you can never get right, if you follow the trend.
Following the trend may lead to more “incremental” results — less exciting and less important
Be a contrarian and work against the trend
someone who is working against the trend. When someone is working on a very hot trendy topic, I’m going to work on something that is not that trendy.
In fact, a recent 2017 paper, titled “BBR: Congestion-based Congestion Control” brought congestion control in the general case back to the spotlight.
So, be a contrarian and working against the trend may be a good idea.
enjoy, trendy topic.
What’s new
In the context of what’s already done in the related work, what’s new in your work?
First, briefly introduce the state-of-the-art in the literature, solving the same problem you wish to solve.
what is the particular state-of-the-art paper that we have?
Then write about what is new in your work that is different from previous work.
talk about what we did, not always talk about what they did.
If the problem is different from previous work — different assumptions, for example — provide justifications why this is important and necessary.
not exactly the same problem?
If the case, you probably cannot directly compare with the state-of-the-art, because problem is slightly different.
you need to do is to provide sufficient justifications on why you slightly different problem is really important, probably it is more practical as compared to the state-of-the-art and really necessary.
This is the first paper talking about solving this particular problem. And if it’s really important, try to justify that.
How is it better?
Now that your ideas are new, they should also lead to better results than before.
Is not just different, but also have better results.
Show how your solution is better than previous work: such as stronger theoretical properties or better experimental results
force more on originality, they talked about “my original contributions,” this is really something that is new. But you cannot be just new, you have to be better as well.
The hard work
Two thing that you must do
Reading and Writing
you also need to read and write, and read and write, in more of a interactive way.
read a little bit, and write a little bit.
Read — Goal: become a expert on the problem
Read — But how?
How do we become an expert in this particular area of research?
Read — Start from one paper
So which paper should we start from?
Read — Perhaps a highly cited seminal paper
in this area of research on this particular topic, sometimes we call it a seminal paper.
maybe they is the sort of like a survey paper, or maybe this is the most recent paper, or randomly downloaded from the web by searching for this particular keyword.
Just start from a particular paper.
Read — Then do an expended-ring search.
Do an expended-ring search from the single paper.
expended-ring is mean the BFS.
Read — Papers that cited this paper
Read — Papers that it cited
go to the references.
Read — Papers authored by the same researchers
Find the papers that are written by exactly the same set of authors.
Go for the papers that are in the same conference, or the same session.
Read — Try not depend only on a search engine
The search engine will give you some random hits, that might be helpful when you get started, but it may not be very helpful because it’s not exactly very comprehensive.
Read — To many, you need to do two thing
One is you need to narrow down on the actual problem that you wish to solve, so all you need to be proficient or be an expert on, would be just one tiny slice of particular research topic.
Just one problem, so there may not be too many problem papers on that very single problem. Very tiny little problem.
But the second possibility of solving the problem of having too many papers is that you need to read quickly.
Read — Read quickly first
You cannot spend many hours reading these papers, you have to spend a very small period of time.
especially if it’s the first time you read this paper. Just grab the paper and read it over a coffee. And that means that you are going to have to read it in 10 minutes, 20 minutes, you should not be spending more time than that to quickly scan through this paper.
Read — Read more carefully if necessary later
And then, of course, if you really like this paper and this paper is really very closely related to what you wanted to do, then you can read more carefully if that is necessary.
That is mean you can spend more time, of course, and that’s okay because now, at this particular moment, you don’t have too many papers.
A lot of papers very quickly, but very small number of papers more carefully.
Give you a lot of depth when you actually read a very small number of papers carefully, and a lot of breadth in the sense that you will never miss a very important paper.
Read — Try to have an interactive way between the two tasks: reading and writing
Write — Write about what you understood
Write anywhere, if you spend 10 min or 20 min or sometime to read a paper.
Write — Write every day
You need to write early, write every day.
You are going to write just two sentences, that’s fine, that’s writing. Two sentences would be good enough.
You have to write every day.
You have to write very very early.
idea → do research → write a paper(two months later, or five months later)? That is wrong.
Writing itself is a part of doing research.
idea → write a paper → do research. The writing process is the process of doing research.
Write — If I don’t have a idea, what do I write about?
There are nothing I can write about.
Write — You write anyway
Write — Write about related work
That you can write because you have been reading papers, right?
You can write about what have you read.
Write — Write about what’s challenging
What’s really hard? Something’s really hard — probably related papers have not even solved completely.
That is a new idea that you are having in you mind when you are reading other people’s paper.
Write — Write about why the problem is important
You need to try to thing about the problem, and say, okay maybe this problem is important, and we wanted to write about it.
You might say, oh, other people have already talked about it!
Well think about it yourself, write about why this problem is important without copying and pasting from the other papers.
In the further, you will see some papers are really mediocre.
Write — Write about what needs to be fixed
That’s something you can write about.
Write — Write a survey paper
So you can write a survey paper, essentially you are basically collecting all of the ideas and you put them all in the paper.
So you’ve been writing a paragraph or two paragraphs a day, and you put them all together a week later or a month later, you have a paper.
So, that is a survey paper that has nothing to do with original research, it is basically just summarizing existing papers. But that would be a paper that you can submit for publication.
In fact, many people did and a lot of these survey papers got very well cited.
Write — Write about what’s new in the context of related work
So, we want to write as we read.
That would be something that goes back to our three essential elements, the problem, what’s new, how is it better?
When we are reading, we’re reading these related papers, and when we’re writing, we’re writing about what’s challenging, what’s really something that needs to be fixed. And now, we come to an original idea, what’s really new, what can we do that is different or better than existing papers.
We have something a little bit more substantial to write about. We have better idea when we actually write about the more mundane ideas of what’s challenging and what might be a problem that needs to be fixed.
So that is the process of creating something new.
Write — Creating ideas interacts with writing closely
Write — Writing a the best way to force yourself to think clearly and be focused on what you really want to do.
When you’re reading, you are just consuming things.
Write — and to crystallize what you don’t quite understand yet.
You use the writing process to crystallize what you don’t quite understand yet.
And that crystallization is very important, you previously have a vague idea if you don’t write, you roughly speaking know what you know about, what you may be able to create, when you start to write, you make it much clearer because you kind of try to crystallize it.
Write — Writing also open a dialogue
A dialogue is something that really is important for the creative process. Because a dialogue is a dialogue with someone else.
Write — For others to read what you wrote
If you don’t write, nobody can read your mind.
If you have written even just two paragraphs, people can read what you wrote, but that would be very important to stimulate discussions.
Write — To stimulate discussions with others
If you don’t really have some kind of a common ground, some kind of a common understanding, with the other people, there’s nothing that you can discuss.
This is actually why you use writing to bootstrap the process, to bootstrap the discussions that you have with your colleagues, with your advisor, with anybody that you know very well, probably people working in the same area, or probably people that may not be working in the same area.
You want to read not only for yourself, but also for opening up a dialogue, so you need to write for your readers.
Write — That’s why writing is a slow and painful process
you’re creating something new, that’s going to be slow.
When I was writing, I’ve basically written the entire five paragraphs, and I will just call it a day, and it took two hours.
When he is writing, he’s actually writing one paragraph at a time. So he’s writing one paragraph. And he is trying to change it, and change it, and tweak it a little bit more, and try to write very very slowly.
So no matter what, it is going to be very very slow.
Use the entire evening.
But no matter what, it's not going to be a fast.
Write — Write in a crystal clear way
由于你是写给你的读者，因此你需要写的一清二楚。
Write — Write about one problem and one (potential) solution
如果你想写一个问题，你需要写清楚你的问题是什么以及认为可能存在的解决方案。
Write — Write with a flow of ideas that’s easy to follow
This is something that is really something that you want to write to work for your readers.
Write — Keep your readers engaged throughout the paper
Slow and painful process
Write — Write your paper so that it’s as easy to understand as absolutely possible
Write — Your readers don’t have to work hard
Write — Write your paper slowly, so that your readers can read quickly
Once we have written some kind of a draft, we can submit the survey paper, we can also try to ask a colleague or advisor to actually try to read what we had written.
The art
So now we have putting the hard work. We have an idea. It might be quite a good idea but how do I actually find a right actual paper. The survey paper anymore this is the actual paper. So we want to pay attention to the title, abstract, introduction, and flow of ideas.
Art — Pay attention to the title, abstract, introduction, and flow of ideas
Different papers are completely different, which means that it’s going to be very tough to talk about the rest of the paper but there are Something that we can talk about today.
不同的文件完全不同，这意味着要讨论文件的其余部分将非常困难，但今天我们可以讨论一些事情。
Title: 2 lines, 10,000 readers
Abstract: 10 lines, 1,000 readers
Introduction: 100 lines, 100 readers
The rest of the paper: 1,000 lines, 10 readers
你需要写的每个部分都足够吸引，以至于读者能够一直读下去。
Title — What’s a good title?
Title — It attracts a reader to read the abstract.
Title — It reflects the essence of the new idea.
Title — Maybe It is as simple as possible
does not have to be a precise summary of the paper.
does not need to include all the keywords.
Title — The best title is one that is the easiest to understand at a glance
Let’s say we have a new idea that uses a new coding technique to reduce the latency when delivering data over the Internet.
Which title is the best?
LLRC: A Low-Latency Rate-Controlled System for Fast Data Delivery over the Internet
Code Information Distribution: New Content-Delivery Protocols for the Internet
Coding Reduces Latency over the Internet
all three titles are good titles.
would personally prefer the 3nd title.
Really gives you a sense of What’s going on. It ‘s Certainly more General it doesn’t talk about You know the specific idea that you have in your paper.
Give the general direction what’s going on.
S.U.Stich. “Local SGD Converges Fast and Communicates Little”, Int’l Conference on Learning Representations (ICLR) 2019.
The title is perfect.
Which title is the best?
Extended Deferred Acceptance: Interference-Free Matching between Spectrum Buyers and Sellers.
