EECS 582
Apr 18, 2013
Prabal Dutta

- Kinds of papers
  - Design and Implementation
    - FFS, TCP, Capriccio
  - Architecture/Framework
    - SEDA, Click, HomeOS
  - Experience/Revealed Truth
    - Hints

  - Vision/Case For
    - TerraSwarm
  - Measurement/Analysis


- Can you:
  - Identify an important problem?
    - How?
  - Demonstrate understand the related work?
    - How?
  - Formulate a crisp hypothesis?
    - What are the elements?
    - Observations, claim about dependent and independent variables
  - Plan and execute a research project?
    - Identify important experiments/analysis
    - Demonstrate step-by-step argument formulation
    - Fix weak links in the argument chain
  - Convey the results crisply, concisely, clearly, and convincingly?
    - Crisply
      - Results must be stated explicitly in the abstract, intro
      - Results must demonstrate some insight
    - Concisely
      - Don't waste words
      - Paper should be as short as possible, but no shorter
      - Must be possible to understand work from abstract, intro, figs, tables
    - Clearly
      - There must be an 'ah ha' graph or set of graphs (or proofs)
      - Every caption must be self-sufficient and descriptive
        - What is the figure?
        - What do the data show?
        - Why does the figure mean (i.e. why should the reader care)?
    - Convincingly
      - Error bounds
      - Confidence intervals
      - Multiple trials
      - CDFs, where appropriate
  - Demonstrate validation of your hypothesis?
    - Does it work?
    - Does it really work?
    - Does it really, really work?
  - Identify the shortcomings of your approach?
    - When does it fail?