My Dear Judge !
When you send the mail with the subject 'New Judgement', please send the old source code that kept you out of sleep (until you take away others).
nirjon.
Please stop this rejudgement rate
Moderator: Board moderators
-
Shahriar Nirjon
- New poster
- Posts: 16
- Joined: Tue Dec 03, 2002 9:44 pm
Hello, i am not speaking about those problems that must be precalculated to give efficiency. I am speaking abt those probs those should be changed to avoid precalculation. I do not disagree to make new problem with increased data set. But will the moderators agree??
And, I agree to the system proposed by Nirjon.
And about website srces, no comment??(better if spoken by Shahriar Manzoor bhai)..
And to sumit bhai, what are the guidelines do you propose?? Would you please mension?
Thanking you all..
Anupam
And, I agree to the system proposed by Nirjon.
And about website srces, no comment??(better if spoken by Shahriar Manzoor bhai)..
And to sumit bhai, what are the guidelines do you propose?? Would you please mension?
Thanking you all..
Anupam
"Everything should be made simple, but not always simpler"
First, I want to say that, for rejudged problem 147 or 357, I still have some ways to use precalculated method to get accepted. I haven't done that yet, but just tell you that you can't prevent people using precalculation easily.
In my opinion, it is not so meaningful to change the problem specification to prevent 0.00 runtime. In a programming contest, there must be easy porblems and difficult problems. Why must you erase the easy one?
And I think that precalculation is not so evil at all. In many practical application, precalculation is a good method to give us fast application programs. And I know that there are some regional contest problems require precalculatied answer.
In my opinion, it is not so meaningful to change the problem specification to prevent 0.00 runtime. In a programming contest, there must be easy porblems and difficult problems. Why must you erase the easy one?
And I think that precalculation is not so evil at all. In many practical application, precalculation is a good method to give us fast application programs. And I know that there are some regional contest problems require precalculatied answer.
My signature:
- Please make discussion about the algorithm BRFORE posting source code.
We can learn much more in discussion than reading source code. - I HATE testing account.
- Don't send me source code for debug.
I agree with a lot of the above posts. Fixing problems with ambiguous/unclear descriptions or fixing incorrect/incomplete data sets is okay since it will improve things. But arbitrarily changing problem statements just frustrates a large number of people for little gain! 
Moreover, I still haven't received notification about my results being changed to WA for a number of problems. (I'm guessing 147, 357, and something else - my count has gone down by 3, but it's not clear what the 3rd "rejudged" problem is). Are there other people who haven't received the (expected) notification e-mail? It would certainly be nice to at least receive a notification e-mail when AC -> WA! (And placing details in the e-mail about the cause of the change, e.g. problem respecified, new time limit, etc. or an link to this relevent information would be even better).
Moreover, I still haven't received notification about my results being changed to WA for a number of problems.
(I'm guessing 147, 357, and something else - my count has gone down by 3, but it's not clear what the 3rd "rejudged" problem is). Are there other people who haven't received the (expected) notification e-mail? It would certainly be nice to at least receive a notification e-mail when AC -> WA! (And placing details in the e-mail about the cause of the change, e.g. problem respecified, new time limit, etc. or an link to this relevent information would be even better).
I just checked that, 357 can be solved by Java 
http://acm.uva.es/cgi-bin/OnlineJudge?S ... :1.00:60::
The program runs for 0:00.348
and my C program using same algorithm runs for 0:00.033, about 10 times slower.
http://acm.uva.es/cgi-bin/OnlineJudge?S ... :1.00:60::
The program runs for 0:00.348
and my C program using same algorithm runs for 0:00.033, about 10 times slower.
My signature:
- Please make discussion about the algorithm BRFORE posting source code.
We can learn much more in discussion than reading source code. - I HATE testing account.
- Don't send me source code for debug.
I think pre-calculation should be allowed (not because I use it though).
Most of the questions available now are pretty simple and we can easily come up with algorithms that hit sub 2 seconds for most of them. But 6/7 years ago when I first tried them, this was not the case. Most algorithms took over 5 seconds and sometimes, the best answer could take quite some time indeed.
In another 5/6 years time, when Pentium 6, 7 or even 8 is out, brute force may even work for questions like 239. Does that mean we change the specs of every question everytime intel decides to release a new line of chips?
I believe that instead of changing old questions every now and then, why not just leave them as they are. Let them serve as archives of what we went through for future programmers to look at. Let them serve as guiding steps for fledgling programmers of the future.
Instead of respecifying old questions, why not come up with new ones that challenge the programmer more?
Right now, many solutions are considered brute force simply because the problem itself is NPcomplete. But as computer science progresses, some NPcomplete questions may turn out to have efficient algorithms. That does not mean we have to change the specs of all such problems. Instead, come up with new challenges that have no simple solutions. That is the way forward.
Most of the questions available now are pretty simple and we can easily come up with algorithms that hit sub 2 seconds for most of them. But 6/7 years ago when I first tried them, this was not the case. Most algorithms took over 5 seconds and sometimes, the best answer could take quite some time indeed.
In another 5/6 years time, when Pentium 6, 7 or even 8 is out, brute force may even work for questions like 239. Does that mean we change the specs of every question everytime intel decides to release a new line of chips?
I believe that instead of changing old questions every now and then, why not just leave them as they are. Let them serve as archives of what we went through for future programmers to look at. Let them serve as guiding steps for fledgling programmers of the future.
Instead of respecifying old questions, why not come up with new ones that challenge the programmer more?
Right now, many solutions are considered brute force simply because the problem itself is NPcomplete. But as computer science progresses, some NPcomplete questions may turn out to have efficient algorithms. That does not mean we have to change the specs of all such problems. Instead, come up with new challenges that have no simple solutions. That is the way forward.
Rejudging - necessary?
I am one of the people that is assisting with the rejudging of many of the problems on the board. However, so far, I am only responsible for solving one of the "unsubmittable" problems and then building a data set for it.
When it comes to rejudging problems, I don't think that a person who solved it before, and now gets a TLE error should be counted incorrectly. I have lost only 2 problems that I know of, and both now have a TLE. For which one of them is "500!"...which makes no sense since my way of doing it was even faster than the way suggested by others.
If a person was storing the data beforehand, and this is the main issue, then just change the memory requirements on it, not the time limit. The suggestion about forming new problems instead of changing the old ones would be good as well. Maybe just correct the old ones if needed, and form a similar problem that is more difficult with a larger data set that cannot be brute forced.
When it comes to rejudging problems, I don't think that a person who solved it before, and now gets a TLE error should be counted incorrectly. I have lost only 2 problems that I know of, and both now have a TLE. For which one of them is "500!"...which makes no sense since my way of doing it was even faster than the way suggested by others.
If a person was storing the data beforehand, and this is the main issue, then just change the memory requirements on it, not the time limit. The suggestion about forming new problems instead of changing the old ones would be good as well. Maybe just correct the old ones if needed, and form a similar problem that is more difficult with a larger data set that cannot be brute forced.
Cheers,
Ron
Ron