Reply to thread

Message: [QUOTE="Joshua Cranmer, post: 3616713"] But this is the dilemma: we have no a priori reason to believe that it's right. You have given no proofs of correctness, whether formal or informal. The burden is not on us to show that you're right, it is on you. I don't trust any algorithm I generate to be right; I have to feed it a few cases that look edgy, and then I have to give my code a test suite to test those edge cases. Also, your algorithms tend to be underspecified making finding code that breaks it difficult, because it's hard to know what will happen next. Code, or pseudocode at the very least, is a powerful way to explain in full specificity how an algorithm operates. So, before you can show that it is right, you essentially have to code it. Resistance to this fact is...pointless. People would complain a lot less if you had code, people would be more inclined to help, and it's not like you're writing anything near the complexity of [looks at latest project] a Java decompiler, so you can't seriously not code just because it's too hard. Translation: I'm a genius, my work is perfect, you're all just too stupid to realize that. Rebuttal: If you think that merely /positing/ that the algorithm is correct will get people to integrate it in their work, you're wrong. TSP is a well-established problem space, that has many "good enough" algorithms. With this in mind, a polynomial-time algorithm would only be of interest if it was correct. So you have to end up showing correctness yourself, which you are surprisingly loath to do... [/QUOTE]

Verification