Monday, April 5, 2010

Learn from Reviewers

Every paper must state (checklist):

   1. the problem addressed (novelty)
   2. the solution proposed (correctness)
   3. an example that shows how it works
   4. an evaluation, ideally a comparison with existing techniques (completeness of experiments)


Title (cover the most important information):
   1. New problem (new perspective)
   2. New findings
   3. New methodology

Abstract:
   1. Problem
   2. Solution or methodology
   3. Results

Introduction:
   1. Why this problem is significant
   2. Recent solutions
   3. What inspires this idea

Methodology:
  1. Theory 
  2. Mathematical formulation
  3. Algorithms
  4. Computational Issues
  5. Implementation issues

Try to package your method within a good theoretic framework ("why to do" is more important than "how to do").

Results & Discussion:
  1. Data sets and performance measures
  2. Experimental settings.
  3. Comparison with state-of-the-art methods
  4. When your method is better and when your method does not work

Posted via email from Troy's posterous

No comments:

Post a Comment

Google+