Confessions of a researchaholic

November 20, 2015

How to have a bad research career by David Patterson

Filed under: Real — liyiwei @ 7:24 pm

The filename suggested the seventh version, but I still learned new things from it just like the previous ones.

David Patterson did warn about his system bias, though.
In particular, I have different experiences with “big teams” and “technology transfer”.

The number of collaborators should be naturally proportional to the scale of the projects, which in turn depends on the fields.
System architecture naturally involves various experts from PL/compiler, OS, architecture, VLSI, and networks, but in my fields (graphics, vision, and HCI) the most influential papers are usually authored by one or two people, and seldom more than four.

It is difficult for companies to adopt new system architecture, but much easier for them to “steal” algorithm ideas. It cannot hurt to protect your IP before going out selling your ideas; you can quickly file provisional patent applications at low cost without involving any lawyers.

November 19, 2015

Research seminars

Filed under: Real — liyiwei @ 1:51 am

To my bright students:

I notice that you do not seem to come to seminars often, if at all.

If you are around, I highly recommend you attend these seminars, at least those by renowned researchers.

You will have fun, learn a lot, and most importantly, have a chance to interact with and talk to different people.

I know you have a lot of works to do, but these can wait, unlike the seminars and the speakers.

October 17, 2015

How to quit jobs

Filed under: Real — liyiwei @ 6:50 pm

[Note: most of my blog posts are answers to common questions, including this one. They do not reflect my personal situations.]

Do not use resignation as leverage for negotiation. If you have any grievance for you job, talk to your managers first. You might be surprised how much you can achieve by simply talking straight.

Quit only if you really mean it, and carry it through. Do not change your mind if people entice you to stay. If you do, they may think you are soft, and you may suffer as a consequence in the long run.

Always attribute your resignation to personal reasons. Do not say anything remotely negative about your employer, managers, colleagues, or underlings. (The only exception is when someone you know gets a job offer from one of your former employers and needs your opinion.) If they insist on getting your honest feedback for future improvements, ask them to contact you after a cool down period. I recommend at least 6 months, and one year if you go to a competitor.

Instead, always try to say something positive about your former employers. If you do this before you quit, you might actually change your mind.
The grass always seems greener on the other side.

Do not come back to the same employer within the next 3 to 5 years. If you do, they might think you are weak, and you might suffer as a consequence.

October 8, 2015

Back to the future

Filed under: Real — liyiwei @ 3:23 am

For those of you who think there is (or will be) nothing new under the sun in your fields:

At some point during the 19th century people thought Newtonian physics is basically done and only incremental works remain.

Then came stuff like relativity, quantum mechanics, etc.

October 6, 2015

Whether and how to contact professors for school applications

Filed under: Real — liyiwei @ 2:07 am

Professors, at least those at top schools, tend to receive many applications, probably hundreds if not thousands every year.

The best way to get their attention is to find people with good reputation and existing relationships as your references.

The best way to get them ignore you is to send out blanket spam-like inquiries to everyone.

Look at the professors’ websites, which may already spell out whether and how to contact them, and other useful information like their research and style.
It is to your benefit to read these, as you want to find someone with matching style and interests, and the professors want to know why they should pick you.

Disclaimer: this is based on my (limited) experiences as a prof so far, not how I contacted my PhD adviser almost 20 years ago.

September 23, 2015


Filed under: Real — liyiwei @ 6:17 pm

If you have to describe your understanding to me to see if you really understand it, then you probably don’t.

The only way to know for sure is to produce concrete results.
This is why I like science and engineering, where things can be measured objectively.

July 27, 2015

Aaron Swartz on how to hire

Filed under: Real — liyiwei @ 9:51 am

This guy knows how to hire, at his early twenties, better than most of my colleagues so far.

July 18, 2015

Write paper introduction to vet research ideas

Filed under: Real — liyiwei @ 11:04 am

The introduction part of top research papers shares the following common structure, which I use to vet and hone potential ideas while drafting the paper/project:

What problem we are trying to solve.
Why it is important, and why people should care.
[Picking the right problem is the most important stage of a research project. Proceed only if this part is very convincing.]

What prior works have done, and why they are not adequate.
Say only high level big ideas. Details should go to a previous work section.
[If you can find only details but not big ideas, it is a sign that the problem domain is saturated.]

What our method can offer: sales pitch for concrete benefits, not technical details.
[Imagine we are doing a Super-Bowl commercial here; every second costs millions of dollars. Lacking of significant benefits is a sign that the idea is incremental.]

Our main idea, giving people a take home message and (if possible) see how clever we are.
[“If you can’t explain it simply, you don’t understand it well enough.” – Albert Einstein]

Our algorithms and methods to show technical contributions and that our solutions are not trivial.
[This part is less important for truly brilliant ideas, but still necessary for the rest majority.]

What are the (anticipated) results, applications, and benefits; stuff ordinary people would care in the real-world.
[Let’s do something really useful.]

July 12, 2015

Fail fast

Filed under: Real — liyiwei @ 2:30 pm

If you are mentoring someone who is clearly not suitable for the profession, and you think you are being nice by keeping him/her as long as possible, you are actually doing a big disservice to everyone because:

. The time and efforts you sink on that individual can be more effectively spent on other tasks, including mentoring other more qualified people.

. That someone can be happier and more successful at doing something else. You are delaying or (worse) depriving his/her opportunity.

. You are degrading the performance and reputation of yourself and your institution.

. That someone can set up a bad example and cast off negative influence to nearby people.

. You are wasting money and resources sponsored by taxpayers or shareholders.

June 25, 2015

How to find and attract top recruits

Filed under: Imaginary,Real — liyiwei @ 9:31 am

Major König: I’ll fix it so that he’s the one who finds me.

Next Page »

Theme: Rubric. Get a free blog at