如何阅读论文——沈向洋院士直播（You Are How You Read）

What is this talk mostly about ?

This talk is mostly about reading.
   ● 如何高效有效的阅读论文
   ● 有哪些工具来帮助阅读

Why reading a research paper？

   ● 导师/同学叫我读
   ● 它在课程所需阅读清单上
   ● 它是正在阅读论文的相关论文

   ● 我需要写关于这篇论文的总结
   ● 我需要在同学前做关于这篇论文的报告
   ● 我需要写篇与该论文相关的调研报告

   ● 我需要仿真这篇论文的部分内容
   ● 我需要为会议或期刊总结这篇论文
   ● 我希望从这篇论文中学习借鉴，为将来的工作准备

Why is reading research papers so hard？

   ● Hard to read：许多文章写的不好
   ● Difficult to understand： 缺乏对其相关topic的了解
   ● Easy to get stuck：当阅读过程中遇到问题，该向谁或者在哪里寻求帮助
   ● What to read next：在读完这篇论文之后，该接着读哪一篇？
   ● Impossible to fucus：在Internet环境下很难花很长时间读一个研究
最最重要的原因是：
The disconnect betweem reading and writing.

Writing is like encoder，and reading is the decoder. Codebook us the knowledge that the writer has and the reader needs to possess(or acquire).

Deep reading leads to deep understanding.
Shallow reading means shallow understanding.

How do you approach reading papers?

1. Quick skimming
   Figure browsing
2. Critical reading
   Creative reading
3. Deep analyzing

   ● Reading well is about understanding how the paper is written.

Paper types?

首先我们要明确我们所研读的论文属于以下三种类型中的哪一个：
   ● Propose/formulate a new problem
   ● Solve a problem peple care about
   ● Exposition and survey

How should we read these papers?

Quick Reading

在弄明白如何阅读之前，我们要先搞清楚如何写论文：
A paper has four parts. Spend equal amout of time writing each：
   1. Title
   2. Abstract
   3. Introduction
   4. The rest of the paper

Quick reading and shallow learing is very important.
Reader cares more about the first two pages(1,2,3)(maybe conclusion).
   ● Is this paper worth reading?
   ● Should i know something about this paper?
   ● Can this be quickly rejected(by conference)?

Write a dynamite introduction:

1. You must make your paper easy to read.
2. You’ve got to make it easy for anyone to tell :
   ● What your paper is about
   ● What problem it solves
   ● Why the problem is interesting
   ● What is really new (and what isn’t)
   ● Why it’s so neat
3. And you must do it up front.

Critical Reading

Ask questions, be suspicious.
如果作者尝试解决某些问题：
   ● 他们是否在解决正确的问题
   ● 有没有哪些情况没有考虑在内
   ● 这个解决方案的局限性在哪里
如果作者展示了数据：
   ● 这些数据是否有利的证实论点
   ● 这些数据是否是通过正确的方式得到的
   ● 他们是否通过某些方式干涉了实验数据
   ● 还有没有更加引入注目的数据集
作者的假设是否合理：
   ● 给定的假设、逻辑是否清晰合理
   ● 推理是否有缺陷

Deep analyzing

Analyze with a list of questions.
以下问题更多的是有关文章的，属于客观问题：
   ● What’s main claim? Key idea?
   ● What’s key limitation?
   ● Is there code available ? Data?
   ● Is the idea neat? Is it counter-intuitive?
   ● Is the contribution significant enough?
   ● Is the experimentation good? Manual tuning?

以下问题属于主观问题，但对于理解同样的重要：
   ● Are there important related papers I missed?
   ● Is it useful to my work e.g. product dev?
   ● Is this a paper worth following?
   ● What do other peple think about this paper?
   ● Who are dominating this area of research?
   ● What question should I ask the authors if i had the chance to meet the authors?

How you should read introduction?

直接整篇阅读体验感非常的糟糕与无聊，但是如果把她进行拆分阅读起来便非常的容易：

Title and levels of summarization

   ● Title(one sentence)
    · TL；DR (few sentences)
   ● Abstract (4-8 sentences)
    · Extended abstract (one page)
   ● Introduction(1.5 pages)
   ● Paper(8 pages, double-column, including 1page of reference)

Best Siggraph titles:
  two words, ending with “-ing”
    ● Plenotic sampling, Poisson matting, Lazy snapping

What do you do when you get stuck?

Creative reading: positive thinking

    ● What are the good idea in this paper
    ● Improvements that might make important pratical differences?
    ● Applications/extensions that the authors might not have thought of?
    ● Can they be generalized further?

If you were going to start doing research from this paper, what next thing would you do?

Write a one-page review

   ● **Your summary：**Key idea in a sentence or two
   ● **Learning from critical reading：**any problem?
   ● **Learning from creative reading：**idea for next year’s conference?
   ● **Your opinion：**do you recommend this paper?

Even share this review with others or do a presentation.
当我们阅读的时候，我们总是倾向于认为论文里说的都是对的，但是当我们给别人来展示的时候就可以通过别人的提问来发现问题，这有助于更好的理解文章。

Tools to help reading/what to do when stuck?

Tips

   ● Write a summary, do a presentation.
   ● Note: bib, summary, key points, figures, citations, structures, styles
   ● Ask: people around you, more experienced researchers, the authors
   ● Read more.Keep reading.
   ● Be patient: reading = building your cogitive model !!!

视频链接

真心的祝福各位小伙伴可以一起在科研的道路上砥砺前行
沈向洋博士5.14直播“You are how you read”