Привет, Codeforces!
В 18.12.2023 17:35 (Московское время) состоится Educational Codeforces Round 160 (Rated for Div. 2).
Продолжается серия образовательных раундов в рамках инициативы Harbour.Space University! Подробности о сотрудничестве Harbour.Space University и Codeforces можно прочитать в посте.
Этот раунд будет рейтинговым для участников с рейтингом менее 2100. Соревнование будет проводиться по немного расширенным правилам ICPC. Штраф за каждую неверную посылку до посылки, являющейся полным решением, равен 10 минутам. После окончания раунда будет период времени длительностью в 12 часов, в течение которого вы можете попробовать взломать абсолютно любое решение (в том числе свое). Причем исходный код будет предоставлен не только для чтения, но и для копирования.
Вам будет предложено 6 или 7 задач на 2 часа. Мы надеемся, что вам они покажутся интересными.
Задачи вместе со мной придумывали и готовили Адилбек adedalic Далабаев, Михаил awoo Пикляев, Максим Neon Мещеряков, Роман Roms Глазов, Артем Ferume Иликаев и Руслан AcidWrongGod Капралов. Также большое спасибо Михаилу MikeMirzayanov Мирзаянову за системы Polygon и Codeforces.
Набор задач частично пересекается с Открытой Олимпиадой КФУ, поэтому если вы участвовали в ней — пожалуйста, воздержитесь от участия в раунде.
Удачи в раунде! Успешных решений!
Также от наших друзей и партнёров из Harbour.Space есть сообщение для вас:
Harbour.Space University в партнерстве с Giga (Unicef) предлагают стипендию для получения степени магистра в сфере Data Science, Computer Science и Front-end Development, а так же опыт работы.
Мы ищем различных кандидатов от junior до mid уровня:
Специалист по работе с данными
- Хорошие знания в области машинного обучения
- Опыт работы с инструментами визуализации данных, такими как matplotlib, ggplot, d3.js., Tableau, которые помогают визуально кодировать данные
- Отличные коммуникативные навыки – невероятно важно описывать результаты технической и нетехнической аудитории
- Сильный опыт разработки программного обеспечения
- Практический опыт работы с инструментами обработки данных
- Способность решать проблемы
- Аналитический ум и отличное деловое чутье
- Высшее образование в области компьютерных наук, инженерии или смежной области приветствуется
Аналитик данных:
- Подготовка данных
- Анализ и изучение данных
- Знание статистики
- Анализ и визуализация данных
- Отчеты и панели индикаторов
- Общение и переписка
- Знание предметной области
- Ориентированность на решение
Front-end разработчик:
- Уверенное понимание HTML, CSS и JavaScript
- Знакомство с фронтэнд фреймворками и инструментами, такими как React или Vue.js.
- Навык решения задач, внимание к деталям и страсть к созданию интуитивно понятных пользовательских интерфейсов имеют важное значение
Full-stack разработчик:
- Интерес и опыт в разработке веб-приложений, информационных продуктов и OpenAPI
- Умение работать с облачными службами развертывания (предпочтительно Azure), конвейером Git и CI/CD, а также процессами развертывания.
- Приветствуется опыт работы с проектами с открытым исходным кодом
- Уверенное знание ML
- Опыт работы с инструментами визуализации данных, такими как matplotlib, ggplot, d3.js, Tableau
- Отличные коммуникативные навыки — важно описывать результаты для технической и нетехнической аудитории
- Большой опыт разработки программного обеспечения
- Практический опыт работы с инструментами обработки данных
- Способность решать задачи
- Аналитический склад ума и отличное деловое чутье
- Степень в области компьютерных наук, инженерии или смежной области приветствуется
Все успешные кандидаты будут иметь право на получение стипендии со стопроцентной оплатой обучения (29 900 евро в год), предоставляемой нашими партнерами.
ОБЯЗАТЕЛЬСТВА КАНДИДАТА
Обучение: 3 часа в день
За год обучения вы завершите 15 модулей (длительность каждого 3 недели). Ежедневная учебная нагрузка составляет 3 часа, плюс домашнее задание, которое нужно выполнить в свободное время.
Работа: 4 часа в день
Погрузитесь в профессиональный мир во время обучения. Вы будете учиться у лучших и с первого дня сможете применять свои новые знания на практике.
ТРЕБОВАНИЯ
- Опыт работы в отрасли
- Международный опыт
- Стремление к обучению
- Устойчивое развитие ключевой момент для вас
- Желание работать на общественную организацию
UPD: Разбор опубликован.
All rounds are created equally.
i wish, i can solved the C no problem in this round ...!
i want to know the end of the story how many problem did you solve ?
i had done problem A is after 1 times trying and problem b is after 3 time trying.... and didn't touch problem C ...cause i had maybe 7 to 10 minute left after solving A & B... and i got maybe -37 ratings after this contest (According my standing on this contest).
sad but i become grey from green in this round...although ratings is nothing i still very sad
no issue brother..please don't be sad...the last div 3 contest (in the past night)...i can't solved the problem A in 1 hours 10 minute and then i ditch the contest...it's hurts me a lot yesterday...but i am okay today...somewhere, i had read..."A master has failed much times that a newbie doesn't even try"...this line inspired me a lot...so don't be sad brother...i hope...one day we will became RED coder...just keep trying brother...!
in fact,i solve 6 problem last night(i never do it before) you do not solve A maybe because you just do not see this kinds of question before. Try to solve more question!
Hope the questions are not confusing.
I'm back with one more post-contest discussion stream
I will discuss the problems I solve in this contest.
and i'll be there to discuss it with you aryanc403
i've been watching your past streams too .
Thanks
get job
I have one :)
UBER SDE2
Thanks for that , would love to participate in that ......hoping for a positive delta this time...
Educational rounds are always exciting, don't know what they surprise.
hope i get positive delta :D
same
For some reason I didn't get the notification that the round would be unrated for me (rating must be between 0 and 2,099), and there is no unrated marking on the registrants list. Is this intended (or am I misremembering)?
Sometimes the educational rounds being hard for me But I feel this one will be interesting ☺️
Hope to become specialist this round!!!
Is it rated?
Yes, it is.
Hoping to perform better than my previous contest.
What's the difference bwtween normal rounds and educational rounds? It really bothers me.
Different scoring type and usually educational rounds are supposed to be "education" so you might encounter more standards.
So educational rounds have less ad-hoc problems and require us to use standard algorithms more?
Not necessarily. It might have more standard but still needs ad hoc and similar. Tbh not too different from a normal . For example last round had ad hoc problems but problem E was standard problem.
USACO bronze annually is clashing with this contest :(
let's make it Cyannnnnn
Hope to become an expert after this round.
Speedforces
C <<<< D
i think there is huge gap between them
I couldn't able to come up with an solution for this problem tried using bit manipulation but got wa on testcase 3 do we have to solve this using any other approach.
you mean problem C?
yes
My solution for problem C
To count how many we have for each power use vector or just ordinary array with 30 elements for each power of 2 that could be entered (0 to 29).
For the first type of operation, just increment that value in the vector.
For the second type of operation, go through from 29 to 0 and subtract until the values are finished(refer to the first vector for how many you have with that power of two) or v which was input has less value than the power of 2 we want to subtract. At the end, check if v is 0 or not, if it is answer is "YES", otherwise it is "NO"
You can refer to my submission here: 237768705
Great approach,thanks for sharing your approach. according to you what will be the rating of this problem
Probably, ~1300, but i'm not sure
Another approach....
So to calculate we can get w from out list or not , i just started traversing each bit of w , if at ith bit is 1 in w then we will look that we have ith bit or not in hashmap.If No then we print NO , If yes then decreament ith bit from our hashmap. if there is still some remain in ith bit then we can push it ahead , like we have m[2] = 3 and we have used it one from it so now we had 2 that is (2^2 + 2^2) we can write this as 2^3 so we can increament m[3] by 1. so in genral we have count extra for ith bit then we can increament m[i+1] by floor(count/2.0).
.
This is uncomfortably true...
How to do E?
Using Flows
I guess, Flows with Demands should satisfy the problem. Just throwing in some virtual nodes for outflow from each node and inflow to each node should do the job. Here, a node is a row/column, and edges are delta-based flows on flipping cells. Note that we know the original sum across rows/columns, and only care about needed deltas through flipping. This delta determines the demands on inflow/outflow virtual nodes.
Agree. I'm about the same.
Use minimum cost maximum flow. Specifically, create a node x[i] for each row i and a node y[j] for each column j.
For each cell in the i-th row and j-th column, create an edge (x[i], y[j]) with a flow of 1. Flow represents filling the cell with 1. If the cell originally contains 1, calculate the cost in advance with a unit cost of -1 to offset the premature calculation cost. If the cell originally contains 0, the unit cost is 1.
Connect the source node to x[i] with an edge of flow a[i] and cost 0. Connect y[j] to the sink node with an edge of flow b[i] and cost 0. If you understand that flow is “1”, this step will be quite natural.
Finally, run the maximum flow algorithm. If the flow is equal to sum(a[i]) = sum(b[i]), it indicates that the assignment is valid, and the answer is the minimum cost; otherwise, it is invalid.
Pretty much the same problem
is rating change delta which shows during contest accurate?
Usually the actual rating change is higher than what is shown in by carrot extension, especially in edu rounds. Unless you get FST
someone will become master then :),congrats
Thanks
How to solve E? Is it related to graphs somehow?
Make a flow network with n+m+2 edges. One is sink, one is source and n vertices for rows and m for columns. Make edges with capacity r[i] to all rows from source and edges of capacity c[j] and cost 0 from columns to sink. Also make edges from row i to column j with capacity 1 and cost 1-a[i][j]. Your answer is minimum cost max flow plus the number of a[i][j] == 1 where the i -> j edge was not taken.
E reduces to this CSES problem: Grid Puzzle II
I couldn't able to come up with an solution for C problem tried using bit manipulation but got wa on testcase 3 do we have to solve this using any other approach.
https://codeforces.net/contest/1913/submission/237823558
Speedforces for someone
Unbalanced Contest,
A,B,C were too easy and D is too hard.
Wrong answer on test 24 in E — Matrix Problem ,, can any one give a test case sorry for my bad English
This test fails
Also this
And this
D is somewhat monotonic stack and dp, I thought so
I tried the formula dp[i] = Sigma(dp[j-1]) with a[j] >= a[i] and j <= i result is the product dp of 2 sequences a[0] ... . a[minimum_index] and a[n-1] ... a[minimum_index] but wrong, do you have any other solution
I just thought so, but didn't come up with a clear solution
.
Unfortunately it was easy :(
D fucking boggled my mind. I thought I had it like, 3 or 4 times, only to fail sample, find out that my method is wrong, come up with new method, rinse and repeat. I wrote everything from monotonic stack to segtree today 🤡
A to C is trash btw, but D made the contest worth giving I think (even though I still cannot get it)
Yup, C isn't that bad but A and B are just trash and the round is unbalanced AF
C is also kindaa repeated idea
How to solve D?
What's the idea behind D
DP on suffixes, maintained with a monotonic stack.
Compute for each index how many arrays you can reach that start with it, ignoring elements before it. You can delete any prefix starting from it that contains no smaller values (the monotonic stack is for this) or delete a prefix immediately after it to reach smaller elements further away. I handled this part with a set but upon further inspection it’s really just a copy of my stack ;)
Could you explain the part "or delete a prefix immediately after it to reach smaller elements further away", it's a bit ambiguous for me.
The first case, where we delete until some index before the next smaller value, does not cover cases where the element immediately after the first one is smaller than it. Consider for example the following case:
4 3 2 1
The array
4 2 1
can be obtained by operating on [2, 3], but since 2 comes after the next smaller value after 4, we would not have counted this case. Arrays where the second element is smaller than the first will be obtained by operating on [2, j], if we want the j-th element (smaller than the first) to become the second. And it's only possible if a_j = min(a_2, ..., a_j). So we maintain the set of possible indices j, and remove elements from the set over time.The problem is quite similar to Indonesian NOI contest here
I think the approach should be quite similar. Assume $$$dp_i$$$ is the number of ways that the array ends with $$$p_i$$$, then we can consider that for a range $$$(l, r)$$$ we can calculate number of ways when $$$p_l$$$ or $$$p_r$$$ is the minimum
For each $$$r$$$, there are two possibilities :
find conditions for all values $$$l < r$$$ where $$$p_l$$$ is the minimum from $$$(l, r)$$$ then $$$dp_l$$$ is a contributing factor to $$$dp_r$$$
find conditions for all values $$$l < r$$$ where $$$p_r$$$ is the minimum from $$$(l, r)$$$. To do this, find an index $$$i$$$ such that $$$i$$$ is the previous smaller element of $$$r$$$. Then $$$dp_{i+1}, dp_{i+2}, ..., dp_{r-1}$$$ are contributing factor of $$$dp_r$$$
For the first case, we can maintain a monotonic stack and maintain the sum of all dp within the stack
For the second case, we can use a monotonic stack to maintain previous smaller elements for each index and to find the sum of the dp we can use a prefix sum of dp
you can do it with DnC on the minimums of the array, split the array on the minima and calculate recursively for the left and the right part and merge those results
Can you explain your approach, what would be the base case and the merging condition?
Thanks for such a neat explanation.
Thanks!
Thanks for explaining the approach.
There is just a correction that:
If both l≠1 and r≠n, like you said we are overcounting the case where we remove every element in [l,r], therefore we must subtract 1 instead of adding it back.
Fixed, thanks!
another speedforces round!
I think Problem D is same as this, Except the constraints are tighter.
The hardest step of problem D is to come up with a correct DP. Optimizing it from $$$\mathcal O(n^2)$$$ to $$$\mathcal O(n)$$$ is easy.
Luckily no one submitted this problem during the contest :(
Speedforces lol
Easiest D ever seen
bro you didnt even solved it XD
I did, just didn't submit
Why would you not submit lol
How to solve E?
mincost maxflow on a bipartite graph
With flow min cost, I didn't manage to create a net during the contest sadly ;d
any hints for D? also can someone try to hack my C?
i think my approach for c would give a complexity of number of elements in the multiset * number of queries of the second type * (some constant) but i am not able to think on a good test case for that
make both 1e5 / 2
I tried ,apparently it passes. No idea why.
Why C failed at test case 5?237765289 Could anyone point out what i missed ?
Try this case
speedforces moment
lol was about to comment that
Today is my birthday, and I wrote this round.lol
I think problem F is quite similar to this problem.
why is C on C when it's quite literally just greedy knapsack and zero idea involved?
Well the proof is not obvious. It only works for set of numbers where all smaller numbers divide each element in the set, for example here it is powers of two. It wouldnt work in cases like {1,3,5}
I would say it's pretty obvious considering it's rated *1200~1300 on clist...
Do you really think people prove it before implementing it? Its just wishful thinking
I don't understand why there is such a large gap in difficulty between C and D.
speedforces
if i had an hour more, i can definitely solve D.knew it was some dp and related with monostack stuff.nice problem anyway
F has appeared before: coi15p2. And no, it wasn't like the idea coincided or something, this is the exact same problem. I mean, problem-setters could just avoided this by literally search their problem up, but they didn't, which I don't really get.
As a result, an unusual number of consistently low-rated contestants AC F, which are solved by like, what, 30 people, LGM included. A person whose skill are around Expert — Candidate Master level should not be able to pull that off in a 2 hours contest.
If enough people complain about this, the contest could probably get unrated, or... since F is solved by so few people, and these guys solved so few problems that it shouldn't matter. I don't know. Personally, I want the contest to get unrated to punish these MF
I'm getting +ve delta so :noo:
It has been said before. Problems will repeat, but the round will not be unrated if not done intentionally.
lol cry about it
Very interesting contest, for me problem " C " was super easy even easier than " B " as I love math problems , Thanks for authors <3
Could you share ur intuition for C? C got a lot of ACs . Everybody figured it out so quickly except me
Sure, we know that pow(2,n) it's binary representation is 100.. ( 1 followed by zeroes )
now for any number X it's binary representation is ones and zeroes for example " 1011 "
we can write it as 1000 + 10 + 1 which all of them are in the form of pow(2,n) .. So, if we subtract the maximum pow(2,n) less that X form X we will reach to X = 0.
in this problem we just add one condition that pow(2,n) must be inserted, also we subtract this number from x until it's frequency in the vector or until X become less that it
Wish you got it
That problem D was just so hard!!!
Can someone please hack my C? It's O(n^2) in some cases and I think it should TLE. But I tried the case of 50000
1 0
queries followed by 500002 50000
queries and it didn't TLE.UPD: I resubmit it to judge it on the new test cases and it passes in 1013 ms. So I may have gotten away with it this time. I'll take it as a learning experience.
The complier understand what you are doing with
while (sum >= p && stock[i]) sum -= p, stock[i]--;
And convert it to a O(1) operation.
I don't think this is true reading the assembly on godbolt
You are right. I carefully checked the assembly on Compiler Explorer and I can see the loop even on -O2.
So maybe because everything is in CPU L1 cache so the loop runs super fast.
Yeah. The loop is also very short, the branch predictor is also probably never failing.
currently trying near $$$48000$$$ add queries, and I don't quite know why it isn't getting hacked (or why the setters even set limits too loose)
same solution different complier
RT https://codeforces.net/contest/1913/submission/237820946
ACC https://codeforces.net/contest/1913/submission/237828694
This is because of the y = 0 case when you use __lg: https://codeforces.net/contest/1913/submission/237832586 This is because the implementation of __lg may be different on the different compilers. Functions starting with __ are not necessarily consistent across compilers.
why this get so much downvotes?? just for D was too hard?
Folks, why in the problem C the greedy approach for the queries of type 2 works? For example, in the normal coin change problem the greedy approach won't work afaik, why here it works? Thank you guys, appreciate!
2^n is greater than the summation of 1+2+4+8+...2^n-1 so you can be greedy about using your smaller powers because they can never contribute to be a bigger one
can anyone say why greedy 0-1 knapsack working for problem C, 0-1 greedy knapsack don't work for all denominations right. did there is any special properties of 2 powers that is making greedy knapsack work?if yes what is that prroperty
binary representation of a number?
how it is related, please elaborate. i saw your sol. you are taking maximum possible sum from coins of maximum denomination, like why maximum possible sum from coins of maximum denomination is related to binary representation
so practically when we make the binary representation in base 2 of a number we always take the biggest power of two that is smaller than our number and make that bit 1 and apply the same procedure but for a smaller number, this works because every number has an unique binary representation
Uh how does rating work? Will I get rating for this round regardless of my score? Someone explain pls
Yes, you always get a new rating for all contests that are rated for you. This contest is rated for anyone under 2100. Considering that you're currently unrated (that's 0 basically), it IS rated for you.
Rating works as follows — for any rated contest that you give, your rating is updated based on your performance. If you perform well, it increases, else, it decreases. Anyway, rating is just one factor to competing, in the long run, the experience results in a net positive anyhow. Have fun! :)
Ayo thanks for the reply. Okay so this is a pretty cool system. I did very badly, but eh it's just for fun. It's actually kinda refreshing to compete in programming just for the heck of it.
Problem C was much easier than i thought TT
What's the worse time complexity of your approach? Isn't it O(m^2)?
It is O(m^2) without compiler optimization.
The compiler understand the loop
while(sum+(1<<i)<=x and aux[i]>0 )sum+=(1<<i), aux[i]--;
and generate a better O(1) code to achieve the same result.
That's cool. I had never known that.
Yes! It's o(m^2) and I've noticed a bit late so i received a hack XD. But I resubmitted with a binary search optimization and got ac too with O(mlogm).
I'd like to see that O(1) code.
Wasn't aware that your code was hacked. Curious what the test case was. I try using
and it does not work
What is your binary search approach?
In this line:
while(sum+(1<<i)<=x and aux[i]>0 )sum+=(1<<i), aux[i]--;
Instead of adding one by one, do a binary search over
lo =0, hi =aux[i]
and check ifsum+mi*(1<<i) <=x
The binary search can probably be replaced with division.
$$$desirable$$$ = $$$(sum - x) / (1 « i)$$$
$$$available$$$ = $$$aux[i]$$$
$$$taken$$$ = $$$min(desirable, available)$$$
$$$sum$$$ -= $$$taken * (1 « i)$$$
my Divide and conquer solution to D:
https://codeforces.net/contest/1913/submission/237844397
good one !
can you explain your approach how the index of element is helping you to construct the ans
as you stored (element,index) in root , what is the contribution of left ans and right ans
correct me if Iam wrong
even I thought of the contribution of minimum index but couldn't proceed:D
for each segment l..r i am getting min value and it's index, let it be (mI). then making a call to its left and right excluding the index. left = l..(mI-1) right = (mI+1)..r
now, there can be two cases, case 1: when we include element with index mI in our required final array: if that is the case answer is simply leftAns*rightAns.
case 2: when we don't include element with index mI, this case is difficult in my implementation:
in this case we have to know whether is there a smaller element than mI on left side and/or right side. thus bools isL and isR
if isL, then we have to remove all left side including mI, then ans will be added is right side. similarly, left side might be added. note: left side and right side includes empty array also, then there can be case where both isL and isR, then empty would be added twice thus remove isL&isR.
Nice One!
Speedforces
Plese explain D
Scroll up
Can anyone understand the code for problem D of tourist? I can't understand what dp_nxt and dp_sub mean.
you need to be true sigma ioi enjoyer to understand this code
Honey, I love you too.
How to solve Problem C using bitset.
Won't we get solutions?
Is there a reliable rating predictor available as of today? This one doesn't seem to be working accurately any more.
carrot extension
https://codeforces.net/contest/1913/submission/237789418
https://codeforces.net/contest/1913/submission/237900686 The same!!! The first one was hacked The second one was accepted؟؟؟؟؟
The hacked case is not considered in the original test cases yet
bruh...you literally submitted it in practice
Hello.Why ratings hasn't come yet?
Rating changes for educational rounds usually take more time.
How much time....it's already more than 20 hrs.
how to solve D ????
What about the editorials and rating changes?
They forgot about them
Can anyone tell when will the rating of this contest release. It's already more than 21hrs...hoping soon.
did they forget to update the ratings? Div 3 contest will start soon. How will the ratings be given if they dont update it today lmao.
When will this contest rating update and where is the editorial
Problem C was terrible for its place.
Please add the editorials.
Since no tutorials have been provided yet (as well as my rating change), here are my short hints of problems D, E, F.
Suppose the minimun element of array is at position $$$p$$$, then any operation across this element will be equivalent to operating on both sides of $$$p$$$ separately.
Code: 237781872
Bipartite graph flow.
Code: 237869255
Just think when we change the character at a certain position, what will happen to a longest palindrome substring centered at another position? Then calculate the contribution to each possible operation.
Code: 238134018
When will the rating of this round be updated? I was hoping to become specialist and then give today's div3 as a rated round.
40 min left for next contest to begin and no change in ratings so far. Uneasy situation.
Problem C
Hey can anyone tell the error in this solution 237781165. I am getting runtime error.
I managed to pin point the issue.
You can see these two submissions:
238078175 Accepted
238078130 Runtime error on test 3
The only difference is that the loop variable
i
beingint
orlong long
I think I figure this out.
The issue of your code is that, when you vector
v
has no element, i.e. for type2
query,b
is0
,v.size()
is0
andv.size() - 1
should be-1
.However,
v.size()
actually returns unsigned integer, sov.size() - 1
becomes really big (underflow) and it pass the for loop conditioni >= 0
. Then accessingv[i]
will cause a segmentaion fault.I am not sure why this doesn't happen to test1 and test2 though,
To solve this bug, you can simply push back a
0
to the vectorv
to make sure that underflow never happened. See 238078855 for added lines.Why are the ratings not out yet?
There is an announcement telling that ratings will be updated after the div.3 contest due to technical reasons. See div.3 post for reference
Is the contest unrated?
check up div3 announcement
Attention!
I received the following message from system: Your solution 237799004 for the problem 1913B significantly coincides with solutions mayank246/237799004, Naytive_Ritninja/237800400. Such a coincidence is a clear rules violation. Note that unintentional leakage is also a violation. For example, do not use ideone.com with the default settings (public access to your code). If you have conclusive evidence that a coincidence has occurred due to the use of a common source published before the competition, write a comment to post about the round with all the details. More information can be found at http://codeforces.net/blog/entry/8790. Such violation of the rules may be the reason for blocking your account or other penalties. In case of repeated violations, your account may be blocked.
As that was the approach used by most of the candidates and the candidate is unknown to me. It's just a matter of coincidence that my code matches with other person. Kindly review it. and give me my rating back. BledDest
Your last div3 submissions are also skipped. Coincidence there too?
Wait what.
The screenshot is taken at 12/21 2a.m.(GMT+8)
Edit : Seems like it's back to normal.
What the difference between "Edufcatinal" with nomal?
Is it easier or it have more rules?