Всем привет!
Скоро, 2 ноября, 12:00 MSK, состоится очередной Codeforces Round #209 для участников Div. 2. Как обычно, Div. 1 участники смогут поучаствовать в этом раунде вне конкурса. Обратите внимание на время старта раунда!
Автором задач являюсь я. Большое спасибо Гере Агапову (Gerald) и Лось Илье (IlyaLos) за помощь в подготовке задач, Марии Беловой (Delinur) за переводы на английский, Михаилу Мирзаянову (MikeMirzayanov) за замечательные системы Codeforces и Polygon.
UPD1: Распределение баллов будет таким: 500, 1000, 1500, 2500, 2500.
UPD2: Поздравляем победителей!
UPD3: Разбор задач
Удачи!
wish all participants a very happy Diwali! :)
I don't know in other countries how is this time for participants, but in our country in this time all of students went to school , and they can't participate in this contest .
Is there a specific reason to take this round in this time ?
I dont know about your time zone but it is saturday here :D
In Muslim Countries like Iran holidays is friday. and saturday is begining of a week.
Ok i learned something new.
is only one day holiday in a week . thats very bad :(
ours is a muslim country,but we also have vacation on saturday along with friday. :)
Organising in two or three different days of week and two different times will help everybody to participate.
Весь раунд промучился над С, так и не решил, но мне все равно понравилось))
Восхитительное системное тестирование! Всегда бы так!
Fastest system testing ever!!!
Fastest system test?!
That was the fastest system testing I've ever seen! Great Job! :D
Waiting for updated ratings. :)
Когда будет разбор?
Я надеюсь, что завтра с утра будет финальная его версия.
woah! this is certainly the fastest system testing ever on Codeforces!!!
EDIT-1: why are the testcases still not available for viewing though?
EDIT-2: now able to view the testcases. thanks!
Если системное тестирование прошло так быстро, то почему все еще нельзя смотреть тесты?
Как решалась B?
Ставишь числа в виде 2 1 4 3 6 5 ... И потом K раз меняешь местами числа в соседних парах, например, для k=2 ответом будет: 1 2 3 4 6 5 ...
Возьмем перестановку n, n-1, ..., 1. Для неё значение рассматриваемой разности равно нулю. Если менять местами элементы с номерами 2i и 2i — 1, разность будет увеличиваться на 2 с каждой заменой. Таким образом можно сделать k замен, и разность станет равной 2k.
Распишем сумму последовательность 2 1 4 3 6 5 ... и заметим, что она ноль. На каждый обмен местами результат изменится на 2 => для изменения на 2k нужно сделать k замен.
Я случайно отправил копию =( Хорошие задачи! Спасибо автору.
А можно по-другому:
На первое место поставить число 1, на второе K+1, остальные числа отсортировать по убыванию. Если аккуратно раскрыть скобочки, видно что отсортированная часть сократится.
Мой код: http://codeforces.com/contest/359/submission/4963564
Upd. Отдельно рассмотреть случай K = 0. Тогда просто отсортируем по убыванию.
Можно было сразу сливать перестановку в вывод:
Если на бумажке проверить аккуратно, то такая последовательность всегда проходит
Great problem set.. Thanks to the author and testers!!!
My skipped solution[submission:4965469] turn out to be accepted when I submit it after the contestsubmission:4967854, testdata for problem D may be too weak.
Thanks for the fast systest! :)
any solution for C ?
can anybody tell me what's wrong with my submission 4967509 for problem C? thanks!
same with you...test case 18
My solution even the answer is same with you.....
it seems a overflow problem?
here is a test case that fails your solution 8 2 5 5 5 5 4 4 2 0
Thanks a lot! I forgot to consider the carry.
Nevermind :)
i don't think
long long
is needed, because the integer is always reducing from 10^9. in terms of the code, its magnitude is always reducing.i found my mistake....if anybody would like, compare WA (4967509) and AC (4968673) yourself!
HINT: there is just one extra character in the AC code than the WA one. :P
you weren't adding the powers before. typo or logical error?
umm, i knew i had to add them, but initially assumed that
=
would do that! so i guess this would fall under logical error.and since i passed the pretests, i didn't recheck it thoroughly, but after i failed the system tests, i did. that's how i found the mistake!
8 2 5 5 5 5 4 4 2 0 this test case fails your solution.
Как решалась D?
Заметим, что каждое число может принадлежать только одному искомому интервалу [l..r]. Далее будем рассматривать каждое из чисел, начиная с наименьших, и искать максимальный интервал, содержащий данное число. Если число уже было отнесену к какому-нибудь интервалу, то его больше рассматривать не нужно, поэтому имеет решение за O(N).
"Каждое число может принадлежать только одному искомому интервалу"
Не всегда: 3 3 3 6 2 2.
Поэтому решение нужно доработать. В частности, мое решение 4995186 очень похоже, только для каждого числа кэшируются все позиции, на которых оно участвовало.
systest is amazingly fast, but the rating update is slow :(
systest is O(0.000001), but the rating update is O(n!!!!!!!) :(
look at this submissions:
4966994 4966935 4966174
i think they are same people!
They are my friends (bad friends :D ) and one of them knows my account password !
you can see submited codes for problems A B C are diffrent !
My subs : 4961619 4962635 4967645
Their subs :4962419 4963602 4966596
I changed my account password ...
In my opinion, it's a weak excuse. You should definitely mention UFO or FBI agents.
And this ones. http://codeforces.com/contest/359/submission/4967310 http://codeforces.com/contest/359/submission/4965227
Don't care.
This isn't first cheat and certainly isn't last on . :)
First time to see a new user care about cheaters !!!!???!!
I am looking forward to updating of my ranking...
I see 2 submissions accept in problems C are same : 4966118 and 4969298 http://codeforces.com/contest/359/submission/4966118 http://codeforces.com/contest/359/submission/4969298
Chill, one of them is in virtual participation.
About problem C, we know that if x1 is large enough and a, m is smaller than N(=1e9+7), and x1 ≡ a (mod N) is satisfied, then gcd(x1,m) mod N = gcd(x1 mod N, m) = gcd(a,m) is good.
But if x1, x2 is large enough and a, b is smaller than N, and x1 ≡ a (mod N), x2 ≡ b (mod N) is satisfied, how to solve gcd(x1,x2) mod N ?
Where is the tutorial?
Почему в задаче B не проходит такое решение. вот псевдокод:
for i = 2n..2+k вывод i
for i = 1..1+k вывод i
При тесте 50000 25000 WA система выдает. Хотя всего 100000 различных цифр и разность сумм равна 2k то есть 50000. И по идее это правильно.
For Problem B, Did anyone else thought of this solution?
I got this solution after scribbling four pages of my book. But don't know why it is right!
My solution is the same as that. But my coding is ugly...
BTW, what book is it? It seems quite helpful.
Let me explain my way of looking at it. I will take an example. n = 2, k = 1. Firstly, my permutation should have 2*n = 4 elements. For now, I do not know anything about it so let's say it is:
a1, a2, a3, a4.
According to the given formula,
(|a1-a2| + |a3-a4|)-(|a1-a2 + a3-a4|) = 2*k
In LHS of the above equation, Let's denote expression inside first parantheses as A1 and second as A2.
Now, 2*k is always even. It means that there are 'k' terms which are being added to A1, and 'k' terms which are being subtracted in A2. One way of looking at it can be:
A1 = C+k
A2 = C-k
where C is some constant.
How can I form an expression such that k gets added in A1 and gets subtracted in A2? By reversing exactly k pairs whose absolute difference is 1!
So, if my sequence was 4 3 2 1.
A1 = (|4-3| + |2-1|) = 2.
A2 = (|4-3 + 2-1|) = 2.
Say I reverse exactly 1 pair, (2, 1) becomes (1, 2).
New sequence = 4 3 1 2.
A1 = (|4-3| + |1-2|) = 2.
A2 = (|4-3 + 1-2|) = 0.
A1-A2 = 2 = 2*(1) = 2*k!
One more example, n = 4, k = 2. I will just give the answer. It is based on the explanation above.
Answer : 8 7 6 5 3 4 1 2.
Notice that two pairs (2, 1) and (4, 3) have been flipped.
General solution: Just form a permutation sorted in descending order and flip 'k' pairs as illustrated above.
Sorry for such a long post. Just wanted to be clear in what I wanted to convey)
Hi, Can you please elabroate on
Now, 2*k is always even. It means that there are 'k' terms which are being added to A1, and 'k' terms which are being subtracted in A2
Regards, Nitin
One way of achieving a 2*k difference in (A1-A2) is to have 'k' terms of value '1' in A1 which are being added, and the same 'k' terms subtracted in A2.
Flipping the pairs of numbers helps in doing this.
I dont know why my problem C isnt working. Hope the editorial comes in fast. Is there a way to see the entire testcase, I mean when I see the test case on which my soln failed it shows — finite numbers and then "..."
Д — чуть ли не первая задача, где автор все-таки заставил юзать sprase table, за что ему отдельное спасибо)
what would be the ans for prob C for this test case
2 6 3 3
and why???? please explain me .............
6 is not a prime :) so invalid
even i wasted a lot of time thinking that a factor of
x
could reduce the denominator and hence would have to be accounted for. then i reread the problem and saw thatx
must be prime. started coding the solution immediately after that! :)Is tutorial coming?
Is O( 2*N (lg N)^2 ) solution passes for problem D??? N=3*10^5 .
Depending on how well written it is, it would be possible for it to pass, but my O(40N) takes 0.35s, so it might just TLE.
EDIT: Mine is
. I forgot about GCD working in
. So yeah, it should pass.
i think my solution is O(N * lg n ^ 2) in best form and O(N * lg N ^ 3) in worst form. (but im not sure!) it get accepted in 1300ms. i Think Your Solution will be accept too! here is my solution : 4967080
http://codeforces.com/contest/359/submission/4979724
This is my submission for D. I picked a number and guessed it as my current gcd. then I did a binary search on my pre-computed segment tree. By this i tried to find a L and R for each position i so that gcd(L to R) = arr[i]. I think it takes O((lgN)^2) and the outer loop of N makes the overall complexity O( 2*N (lg N)^2 ). But it is getting TLE on case 10 where N=3*10^5. Is it a problem with my implementation.
That's just what I was talking about above, that a well-written solution would pass. In this case, there's an additional factor of
to the complexity, so a segment tree, which has a large constant compared to some other approaches (sparse table for example), can get TLE.
Seeing as it takes 1.5s on case 9 with N = 2·105, it probably just needs some small optimization to pass.
Thanx a lot Xellos & MiRZA(I don't know how to tag a nick:( )for your gr8 help. Finally Got AC! Same code just replaced the Segment tree portion with sparse Table. Totally new experience for DS beginner like me. Time Limit was 1341 ms, and i am happy with it. Thanx again.
"Tagging a nick": When you write a comment, there's an icon on the top panel (yellow, blue and red stripes). Click on "User" there.
Your Welcome ;)
Wow! All problems by one person!