I am developing an algorithm, and I want to compare its performance with various existing state-of-the art methods. I would like to contact the authors of these methods and ask if they can send me the code for their methods.

What should I tell them? Should I say that I am developing a method that is in some way an alternative to their method? Or is enough to ask to say that I want to do generic comparisons?

  • Is it correct to assume you are not talking about purely theoretical papers? – Bitwise Apr 25 '16 at 14:15

Tell them what you told us. You are working on a different algorithm and implementation for the same problem and you'd like to compare to the state of the art. You've read their papers (cite them), and you'd like to see if their code is available for cross-comparison.

If they don't make theirs available, you may have to reimplement their algorithm in your code as well. This latter case is pretty common. There's a chance you might not do this as well as them, so there's some incentive for them to give you their code to get a proper comparison done.

| improve this answer | |
  • 1
    On the other side of the spectrum, some authors have incentive to not make their code available, as they are looking to market it. – Etheryte Apr 25 '16 at 16:48

You will want to show mathematically that your algorithm performs better than existing methods.

Before you contact anyone, learn a bit about Big-O notation. The real comparison won't be running benchmarks, it will be analysis of your algorithm and theirs. (The other authors have probably already done their part by publishing analyses of their algorithms, or someone else has.) So, you probably don't need the code.

| improve this answer | |
  • 3
    Actually I am interested both in performance and accuracy. I already know theorethical complexity of my algorithm, and also the one of the other algorithms can be found in their papers, and they are similar. My point is not to proof that mine is better, but to understand in which conditions (if any) my algorithm improves and in which it is worse, so it is interesting to have an idea, on the same data, of the difference between methods. – Ulderique Demoitre Apr 25 '16 at 12:01

Not the answer you're looking for? Browse other questions tagged or ask your own question.