Scope of Research
For PhD level, contribution expected at global/ world level
What is Research
- Quantitative
- Def: use of mathematical models, softwares do statistical, formulaic or numerical analysis
- Approach: discovery, analysis, causal determination, prediction: or generalization of findings
- Results: “This solution is N % better”
- Qualitative
- Def: use of non-numeric techniques
- Approach: discovery, analogy, understanding, extrapolation of similar circumstances
- Results: “This is a new problem and here is how to solve it”
- Pragmatic method & Mixed methods:
- Generally problem given
- Combines both quantitative and qualitative approaches
- Advocacy/ participatory research methods
Approaches to CS Research
- Attack a known problem, find or compare tools to solve it
- Formalize a new problem, then attack it
Why Do We Cite Papers
- Relate our problem to prior work
- Add credibility
- Show that we know what is important in prior work, what is solved and what is open.
- Establish current state-of-the-art
READ HIGH QUALITY PAPER
- Tools
- Google scholar
- Microsoft academic research
- Library
- Research gate
- Criteria
- Acceptance rate (for conference/…)
- Total citations
- Impact factors
- Number of readers
How To Read A Paper
- The browse (total 1-2 mins “gut check”)
- Go through 100-1000 papers a year
- To see interesting /relevant
- Abstract + figures + captions
- Venuer check. Author Check. Citation check.
- The skim (5-10 mins max)
- Believe the author and assume it’s all true and decide if the paper is actually interesting and worth more exploration
- Answer:
- Category: type of paper
- Context: what other papers is it related to? How related to your work?
- Contribution
- Credible
- Care: will anyone care, given its valid? will it change anything?
- Cost: time and effort cost you to read it carefully
- 2-3 sencentences to say the addressed problem/ the new in solution/ how significant the experimentation/theory
- Critical read (1 hr)
- Read carefully, but ignore details (e.g., proofs)
- Note assumptions and question them
- Analyze experiments/conclusions
- Anslyze figures, diagrams, illustrations, graphs
- Mark relevant unread references
- Summary the main thrust and point out likely weakness in assumptions/methodology/experiments
- Critically read
- Understand
- The problem
- The proposed solution
- The competing approaches/designs
- The evaluation methodology
- The comparison to state-of-the-art
- Answer questions:
- Is the problem carefully stated/ formulated?
- Is it a meaningful/ real problem?
- Is it the RIGHT problem (formulation vs description)?
- Make a list of assumptions?
- Novel about the problem formulation, if any
- What well known problems are related or the same?
- Are there simple solutions the authors do not seem to have considered?
- Consider every divide, matrix op, function inversion, minimization (maximization) makes sense to you? Are assumptions justified?
- What are the limitations of the soluation?
- Is the logic of the paper clear and justifiable, given the assumptions?
- Do all the pieces of their work fit together logically?
- Has the right theorem been proven?
- Are theorems logically supported? Are proofs given? Cited? If cited, how credible a source
- If data is presented, did they gather the right data to substantiate their argument and did the data appear in the correct manner?
- Did they have enough data to make a statistically sound decision?
- Did they interpret the data in a reasonable manner?
- Would other data be more compelling?
- Did they compare with actual state-of-the-art performance?
- What were the results? Did they do what they set out to do?
- On what dimension did they advance the art?
- Have the authors been cutting corners?
- Would results be reproducable?
- Problematic experimental setup?
- Confounding factors
- Unrealistic or artificial benchmarks?
- Methodological misunderstanding?
- Do the numbers add up?
- Are the generalizations valid?
- Are the claims modest enough?
- Creatively read
- What is the novelty of the paper?
- Does it open up new directions?
- What are the good ideas in this paper? Can they can extended or generalized?
- Could their ideas be combined with other approaches / your work?
- If there were other assumptions made, could that improve the approach? Could you ensure those assumptions were true for some problems?
- Are there better numerical methods than what they are doing?
- Are there possible improvements that make important practical differences?
- If you were going to start doing research from this paper, what would be the next thing you would do?
Literatrue comprehension
- Active reading
- Literature map: https://tobloef.com/text2mindmap/
- write the title of your research on top
- main topics relavent to your research underneath
- Associate the papers you read with each of the topics