Timeline for Loopholes that are forbidden by default
Current License: CC BY-SA 3.0
11 events
when toggle format | what | by | license | comment | |
---|---|---|---|---|---|
Nov 17, 2014 at 23:35 | comment | added | Mark Adler | I guess that's better. | |
Nov 17, 2014 at 23:34 | comment | added | Mark Adler | By the way, detecting an optimization for the minority of cases may be extremely subtle. For example, simple choices in the ordering of what is looked at or the numbering of symbols could be optimized to improve performance on the test cases. Again having no impact on the average performance. This is fine loophole to try to close, but be prepared for ambiguity and arguments about whether it applies or not. | |
Nov 17, 2014 at 23:32 | comment | added | Martin Ender | @MarkAdler how about now? | |
Nov 17, 2014 at 23:32 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 7 characters in body
|
Nov 17, 2014 at 23:30 | comment | added | Mark Adler | With respect to: "but performing very badly for general input", the example that inspired trying to close this loophole permits performing optimally on all cases, yet still permit optimizing for the test cases. In that example, the result is effectively an integer between 0 and 6,670,903,752,021,072,936,959. It is easy to map n test cases to 0.._n_, resulting in compression to a single character (or less) for those cases, while having exactly zero impact on the average performance. So you may want to remove that condition to closing the loophole. | |
Nov 17, 2014 at 20:15 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 15 characters in body
|
Nov 17, 2014 at 20:05 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 303 characters in body
|
Nov 17, 2014 at 18:59 | comment | added | Martin Ender | @abligh Yes, larger sets test cases help mitigate that, but even those can be hardcoded. And scoring by worst-case input only really works for fixed-size problems. | |
Nov 17, 2014 at 18:57 | comment | added | abligh | +1, but also I think it would be useful guidance for question setters to either score on the worst-case input, or take such a wide variety of input that it's impossible to optimise. This is because it's hard to see from an answer whether it's been optimised for the test cases or 'happens' to give a good answer for a small amount of test input data. | |
S Nov 17, 2014 at 18:50 | history | answered | Martin Ender | CC BY-SA 3.0 | |
S Nov 17, 2014 at 18:50 | history | made wiki | Post Made Community Wiki by Martin Ender |