Well, I've already seen in ACM-ICPC people that used an exactly and correct brute-force algorithm for small test cases, and for the big ones the solution wouldn't run on time. But big inputs normally are generated automatically and most of them doesn't contain edge cases. So, sometimes an heuristic can work for big but doesn't work for small, and a team passed a problem using this approach. I saw in GCJ this year too a person that got an incorrect for the small and a Correct for the Big.
On Sat, Sep 19, 2009 at 5:54 PM, Bartholomew Furrow <[email protected]>wrote: > This conversation reminds me a lot of one we had at our last team meeting. > The outcome was this: that we want to make it harder, but not impossible, > to download the Large when you haven't solved the Small; and that we don't > want to change the user interface in the middle of this year's competition, > because it could easily cause more confusion than it saves. So it's a great > idea, and we have a couple of possible UIs in mind, but we're going to save > them for later. > > > On Sat, Sep 19, 2009 at 10:52 AM, Carlos Guia <[email protected]> wrote: > >> In my opinion, not knowing if you solved the small correctly hardly sounds >> acceptable, that information can be found in 3 different places >> - The submission box >> - The scoreboard >> - The "your last ..." message on top of the pages >> >> I understand pressure makes us unable to think right, but this contests do >> measure how you behave under pressure. >> >> Also think about this situation: there are only about 5 minutes left of >> competition when you finish a problem, solving the small won't get into >> qualifying positions but the large will, you think you have the right >> solution, you have to go for the large now, it's your only hope. >> >> Carlos Guía >> >> >> >> On Sat, Sep 19, 2009 at 12:17 PM, Blub <[email protected]> wrote: >> >>> >>> >>> Yeah, having a solution for the small input does not mean it also >>> works for the large input. >>> I agree that this is very similar to real live and don't oppose >>> against making it part of the contest to >>> let the competitors estimate the usability of their large input code >>> without beeing able to test ist. >>> >>> However, in my case, the code at first did not even work for the small >>> input. I overlooked that I failed >>> on the small input and tried the same code on the large input. Later >>> on, when I realized that my code was >>> buggy, I wasn't able to resubmit the large input file. >>> >>> Unless it should be also part of the contest to avoid such stupid >>> mistakes, I see two solutions to improve the situation: >>> >>> 1. A warning, when you try to submit code/output file for the large >>> input when you havn't solved the small input first. >>> 2. Entirely revoking the possibility to submit for the large input >>> unless you have solved the small input first. >>> >>> >>> >>> >>> >>> On 15 Sep., 12:17, rahul chhabra <[email protected]> wrote: >>> > One more point, I have previously read in this group mail is:- >>> > Our applications if they are working for small data set does not mean >>> > that they 100 % also work for large data set. >>> > So, this is exactly simulation of real-life scenarios. >>> > If you don't take care, application is gone. >>> > >>> > On Tue, Sep 15, 2009 at 12:46 PM, ulzha <[email protected]> >>> wrote: >>> > >>> > > The idea of not allowing resubmit is exactly that - are you or are >>> you >>> > > not able to write correct programs without being assessed by an >>> > > oracle. >>> > >>> > > In real life, you do not want to throw together a quick and elegant >>> > > system and deploy it to hackers for intrusion. You do want to cover >>> > > each and every breach by QA already. >>> >>> >> >> >> > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "google-codejam" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/google-code?hl=en -~----------~----~----~----~------~----~------~--~---
