How to do High Quality Research and Run a large Research Group

原创 2011年01月08日 15:45:00

Update: Slide is available at here

 

Yesterday, Prof. C.-C. Jay Kuo from USC gave an interesting talk titled "How to do High Quality Research and Run a large Research Group". This is a talk about philosophy and management, and I summarize some key points as follows.

 


Why PhD?

Prof. Kuo said there are two forces to drive a person to pursue a PhD, one is external drive (family, peer pressure, small success), and the other is internal drive. Research interests is one important internal drive, and a student must of curiosity so as to have passion in research. Strong ambition is also an important internal drive.

 

Problem selection? 

A new problem which is practically important yet technically challenging is the best problem to investigate, but in reality, it is quite difficult to find such a problem due to the limitation of experience, vision, etc.

 

Aim high, accept low?

A student must set a high goal for himself so as to motivate himself. However, reality is always tough. If we come across some failures (rejections...), we need to accept low, and stand up to fight again!

 

Literature survey?

Background survey (tutorial + references) and forward survey (cited function in Google scholar) are two good ways to dig into a new research areas. In addition, when we read a paper, it is very important to have a proactive attitude. That is, after reading abstract and introduction, we should close the paper and ask: if I am the author, how will I write the paper? Critical thinking is very important, or else you will be controlled by authors' logic.

 

Nurturing good taste?

We must read papers from top conferences and top journals. In addition, it is good to divide papers into several categories (I still need to improve my ability in this part):

A: 80% understanding (introduction, results, methods)

B: 50% understanding (introduction, results)

C: 20% understanding (introduction)

 

Writing?

Paper writing is not a trivial task, a suggested solution is multi-pass writing:

1st pass: detailed outline

2nd pass: rapid writing

3rd pass: fine-tuning

4th pass: cross-reading

 

Weekly report system?

Write a weekly report to supervisors and receive their comments could help students make progress in every week. A sample format:

 

1. Tasks finished this week?

2. Plan for the coming week?

3. Feedback and Interaction

4. Detail report

5. Milestone (conference, journals, etc.)

 

 

Doing research is a touch yet joyful journey, keep up Chao, you can do the best!

 

版权声明:本文为博主原创文章,未经博主允许不得转载。

相关文章推荐

How To Do Research In the MIT AI Lab

Introduction What is this?  There's no guaranteed recipe for success at research. This doc...

how to do great research && Writing Tips for Academics

The Paper Reviewing Process Posted: October 18, 2013 | Author: Nick Feamster | Filed under: adv...

MIT How to do Research

  • 2013-02-25 21:28
  • 159KB
  • 下载

How to do research in MIT.pdf )

  • 2013-04-17 09:36
  • 1.91MB
  • 下载

How to do research?

Curiosity (好奇心),organization (组织能力)和meticulousness (谨小慎微),这三个特征是researcher应该具备的。如果你尝试开始一个research pr...
  • gcyxf
  • gcyxf
  • 2015-04-16 17:59
  • 423

How To Do Research In the MIT AI Lab_1_Reading AI

How To Do Research In the MIT AI LabMany researchers spend more than half their time reading. You ca...
内容举报
返回顶部
收藏助手
不良信息举报
您举报文章:深度学习:神经网络中的前向传播和反向传播算法推导
举报原因:
原因补充:

(最多只允许输入30个字)