Overview
It is very important to critique a paper or a talk, understand its shortcomings, suggest future improvements. Thence, incremental research occurs.
General loop: Find area. Find problem in a given area. Find solution strategy for a given problem. Implement solution, evaluate, communicate.
Popular research mode: Knowing and heavily using old techniques: take a few hammers, hit everything with it. Finding new techniques is rare and hard.
OFTEN, perturb the problem and extrapolate a solution. This is a very important skill.
Papers
In doing incremental-improvement research, it is important to be able to critique papers, find their shortcomings, identify unsolved problems, then to be able to solve them.
Talking to others
Another person may be facing a problem in research where you, having a different perspective, may have a good solution strategy. Research shows that this mode of finding and solving problems is results in very good quality research.
Choosing problems to pursue
Judging impact: theory
The most applicable of all is a good theory.
Importance of selectivity
You don’t have enough time and energy to pursue all interests. So, must be selective.
Fresh problems: Importance
Find your own, fresh problems and work on them. If a senior person is not actively involved, it is much harder to beat others who have worked in the field for years.
High vs Low impact work: balance
Aim for rough water. Devote adequate time to important research, like Leslie Valiant.
Short term/ competitive research should be avoided. As Dijkstra [EWD637] said, do not compete with your colleagues who are as well equipped as you are to tackle a certain problem.
Andy Yao said this: When you get a paper accepted to a good conference, the main effect is that it buys you some time: some number of weeks or months during which the pressure to publish is off and you can focus on exciting but very hard open problems. Thus, your work should be organized to achieve a balance between projects that will probably improve your vita, and projects that will most likely not get you anywhere in the short term, but that are important in the long term.
Solving problems
See problem solving strategy.
Parallelism
Execute many research projects/ threads in parallel. Switch between problems. Know when a project isn’t right for the present moment.
Time multiplexing
Must have a good way of dividing time amongst various activities.
This worked well in Fall 2009: Experiments from 4 or 5, reading and classes before that.