Hello Codeforces!

On Mar/22/2019 18:05 (Moscow time) Educational Codeforces Round 62 (Rated for Div. 2) will start.

Series of Educational Rounds continue being held as Harbour.Space University initiative! You can read the details about the cooperation between Harbour.Space University and Codeforces in the blog post.

This round will be **rated for the participants with rating lower than 2100**. It will be held on extended ICPC rules. The penalty for each incorrect submission until the submission with a full solution is 10 minutes. After the end of the contest you will have 12 hours to hack any solution you want. You will have access to copy any solution and test it locally.

You will be given **7 problems** and **2 hours** to solve them.

The problems were invented and prepared by Roman Roms Glazov, Adilbek adedalic Dalabaev, Vladimir Vovuh Petrov, Ivan BledDest Androsov, Maksim Ne0n25 Mescheryakov and me. Also huge thanks to Mike MikeMirzayanov Mirzayanov for great systems Polygon and Codeforces.

Good luck to all participants!

Our friends at Harbour.Space also have a message for you:

Hello Codeforces!

We want to remind you about the two fully funded scholarships we currently have available:

Master’s in Data Science Scholarship & Master’s in Robotics Scholarship

Both scholarship opportunities include: - Full coverage of the Programme’s tuition fee (€23,000 value) - 3 hours of study a day at Harbour.Space University - 4 hours of internship a day with one of our industrial partners - €12,000 euros a year (living allowance)

If you’re interested in applying for the Robotics Scholarship, apply here.

If you want to apply for the Data Science Scholarship, fill out the form below and we will contact you about the next steps.

If you need more information about either, please contact us at hello@harbour.space

Congratulations to the winners:

Rank | Competitor | Problems Solved | Penalty |
---|---|---|---|

1 | QDEZ604 | 7 | 250 |

2 | dotorya | 7 | 254 |

3 | 300iq | 7 | 586 |

4 | dreamoon_love_AA | 6 | 165 |

5 | Hazyknight | 6 | 169 |

Congratulations to the best hackers:

Rank | Competitor | Hack Count |
---|---|---|

1 | Haunted_Cpp | 8 |

2 | 7qk | 4 |

3 | Jobaidul | 3 |

4 | Abu_Musa_99 | 3 |

5 | aihay | 2 |

And finally people who were the first to solve each problem:

Problem | Competitor | Penalty |
---|---|---|

A | Sonechko | 0:01 |

B | usertab34 | 0:04 |

C | KhaleD_ | 0:04 |

D | edisonhello | 0:02 |

E | Roundgod | 0:23 |

F | QDEZ604 | 0:38 |

G | 300iq | 0:44 |

UPD: Editorial is out

It's raining contests

Good to have a rain after a drought

I believe "aloo" flourishes in rain so good for you boi...

P.S. For advanced species : aloo = potato xD

Auto comment: topic has been updated by PikMike (previous revision, new revision, compare).Expecting this round to be full of awesome questions with strong pretests!!!

Thank you, Harbour.Space University for the Educational Rounds. Best of luck everyone :)

Happy CodingI think Codeforces community doesn't like wishes and greets...

I think so :|

There are some idiots also, who put down vote in every comment after a bad contest.

A contest by Vovuh after such a long time.

ROBOTICS a.k.a. the biggest lie and way to tell yourself “i’m good i’m good” and live your whole life in a lie. listen kids, no one cares about your utter trash and useless robots because you’re not the next einstein and the world has enough smart vacuums, smart fridges, smart washing machines etc. (which are better in every way than your “creations” could ever cease to be.)

it’s literally “santa’s real” for big kids / grown adults.

May the force be with us! Happy Coding :)

Is it rated?

its rated for you

Educational Codeforces Round 62 [Rated for Div. 2], Isn't just the title enough to answer your question ?.

there is no "*" mark before the handles of those who are not in DIV2! So is there a change in displaying people who are participating out of contest or only I am seeing this???

There is no out of contest for Educational Rounds. They are for everyone. Just, it is rated for the division 2 participants.

Im running back home for this contest. Wish I would make it home before start...

WOW! MADE IT!!

Contest has started.

My mind:donkey mode has been activated

I want to add some scores.^_^

everybody let us go!

Hope for no 502 bad gateway like there was 3 minutes ago

de ce coaie

Whoever decided the problem difficulties is a troll.

I think the order of Problem C and Problem D should change.

Lmao, order of BCD should change.

That is also right to some extent.

I sense weak test cases in problem D as the percentage of successful submissions is too high

you are right for once, but IMO, they were all easy(A, B, B, B, E)

Also, listen to my new song

You're wrong as always. It was B B C A D

while loop = div2B LMAO

C is just a sorting lmao B is just implementation maybe u r right for D and E, but D is harder than div2a

Problem statements for problems A and B could be written better. Took me a lot more time than I care to admit to understand the problems.

What is the test case 2 for Problem B?

Probably <<>> Ans = 2

I don't know what test 2 is, but the string is good if and only if it starts with a '>' or ends with a '<'. From here on, it should be clear how to solve it.

How's the answer for <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> [2nd test case, 5th one] 50? I could choose the last < repeatedly until all characters before it are deleted, then choose the first > repeatedly until all characters after it are deleted. If you do that, the answer is 1. The problem statement doesn't place a restriction on the number of times you can choose a character, it only specifies that it must exist in the string. If repeated choices are not allowed, then how does >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> evaluate to 0?

You're allowed to remove characters only before you start performing operations on the string, so you can't reduce it first and then remove.

Ahh! Now it makes sense!! Thanks a lot!

True. The statement for A ruined the contest for me.

After 4 minutes I went to read and solve B, then I came back to A. I don't know why, but reading the problem later helped me understand it.

Yes even I felt doomed

Lmao, felt the same D:

Again really nice C it's always fun solving problems that needs thinking

What is solution for E?? What is test 5?

How do we solve E?

Note that there are no odd palindrome substrings of length > 1 if and only if there are no palindrome substrings of length 3. Then, note that a substring is a palindrome of length 3 if and only if the first character equals the last character in the substring. Thus, you can decompose the array into odd and even index arrays, and the problem becomes how many ways to fill in the array such that no two consecutive elements are equal. Then, we notice that we can further decompose each array into arrays of the form "# -1 -1 -1 .... -1 -1 -1 #", where # is either nothing or a positive element. From here, figuring out the number of constructions is an old HackerRank problem, with a solution on GeeksForGeeks.

Yeah, you are right. That problem in Hackerrank was mine. :)

Can u please provide some example or proof for your statement "_Note that there are no odd palindrome substrings of length > 1 if and only if there are no palindrome substrings of length 3_".

Assume there is an odd palindrome of length >1. Then the middle 3 characters are a palindrome of length 3.

Assume there is a palindrome of length 3. Then obviously there is an odd palindrome.

Thanks. Got it now :)

My approach was all the elements at odd indexes must be unique and same for even indexes, what's wrong with this approach?

Well, it's not that they must all be unique, it's that they must not have consecutive elements that are equal.

For example, the array [1, 2, 2, 1, 1] is good, even though your approach would say it is not.

Got it, thanks!

Note that "no palindrome of odd length=no palindrome of length three", then solve for odd/even indices of the array separately.

My idea is not so neat as the others, but I think it works: first I check with rolling hash is the existing array is good or bad. If existing array is good, then you can place any number in place of ? if two ? are at least two characters apart. For the ? which are less than 2 characters apart, we do some case handling. Next we need to check if something like this exists

1???1, or 21??12

So, we then handle such cases, and it's O(n). Think of dividing the array into subarrays delimited by ? on each side of it. Then each subarray contributes to the computation for ? on either side of it.

And finally, we have ???? itself, which is not that hard.

Overall, it's not that hard but for me and for my idea, I can't code this fast enough.

You can see my dp solution here 51717191

What are your states?

I make a dp array with dp[i][k] is the ans for a sub-array with i numbers -1 with and k = 1 if two number at two head are equal and k = 0 if they're not.

How to solve Problem B and C.

You have to find first '>' from left and '<' from end. From this points you can get delete rest of them.

Why? Can you please explain the question and how did you get this. Because I think I am missing something in the question.

From '>' you can delete everything left to it and from '<' — right ones.So, you have delete characters until '>' is first or '<' is last

don't like to criticise individuals, but whoever wrote the statement for problem B is not deserving one. It says " for each operation you have to choose some character that still remains in the string", according to this, one may think that we can chose same character any number of times and then result will either be 1 or 0. But I don't know what he/she intended to write. Disappointed from this round! Also problem A was poorly written too. Took me 25 minutes to understand what the author wants.

That's what I thought in Problem B. Got WA on testcase 2. Can't understand how the problem statement can mean anything else other than the logic which gives 1 and 0 as answer.

You CAN choose the same character any number of times, but the answer is NOT always 1 or 0 :)

Give a counter case

<<>>>>>>>>>>>

Answer here is 2: delete the fist 2 characters, then keep selecting the first character however many times you need :)

Gotcha must read problem statements more carefully

Also I submitted C after the contest using priority queue and sorting is there some other approach also. 51723054

For <<>> if i choose 2nd character('<') then the string will become <>> as choosing < will delete its preceding character. Now i can choose the first occurence of '>' which will delete the last character. The string becomes <>. Now delete any one of the characters will make it a good string. So there's only one deletion. How the ans is 2 and not 1? Please correct me if I am wrong

[copied from above, since this thread seems to have more activity] How's the answer for <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> [2nd test case, 5th one] 50? I could choose the last < repeatedly until all characters before it are deleted, then choose the first > repeatedly until all characters after it are deleted. If you do that, the answer is 1. The problem statement doesn't place a restriction on the number of times you can choose a character, it only specifies that it must exist in the string. If repeated choices are not allowed, then how does >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> evaluate to 0?

From nishank.suresh You're allowed to remove characters only before you start performing operations on the string, so you can't reduce it first and then remove.

It's even in boldface lol. Can't believe I missed that.

LOL XD I also missed it. Thanks bro

It seems dotorya was hacked and 0ahmad0makki0 copied dotorya submissons.

After I finished the contest, I had some submissions on my account which I didn't do. It was almost same with my previous wrong submissions. So, I strongly think that my account was hacked. Now I changed the passwords :(

He though that he would increase your penalty so he becomes 2nd hahahah , that's so hilarious.

He should be banned

It seems it's okay to remove "It seems" in your comment. Other interpretations seem impossible.

you are a seempleton

https://codeforces.com/contest/1140/submission/51701961 https://codeforces.com/contest/1140/submission/51721098

He has also copied your submission.

Edit: seems it was after contest...

but how can one came to know dotorya password

https://atcoder.jp/contests/agc032/standings

Now it's Makki0

UPD:

Now on CodeChef — @a_m_h_m

https://www.codechef.com/viewsolution/23749415

Still on CodeChef, shared #1 in CC long: https://www.codechef.com/rankings/APRIL19A

What is the TC #16 of Problem E?

The

problem Cis interesting ! can't wait the Editorial... any hint ?If you are told the minimum beauty value of your elements, could you easily find out what the optimal elements to pick would be?

Yes ! sort the Length and pick at most $$$M$$$ songs such that the beauty of the selected songs greater than or equal to the minimum beauty i told. right ?

Yes exactly, you would just pick the greatest $$$k$$$ lengths such that the beauty is greater than or equal to the minimum beauty.

If you sort the songs according to beauty, then you can preprocess these $$$k$$$ lengths as juckter said below using a heap, and then iterate over all the possible minimum beauties and take the maximum answer.

could you please explain this part of your code for problem E ...

Sure. dp[i][0] is the number of constructions for the first i elements such that the i’th element is not the same as v.back, and dp[i][1] is the number such that it IS equal to v.back.

That is the base case, so if the first element is equal to the last, dp[0][0] will be 0 and dp[0][1] will be 1, and vice versa if they are not equal.

Awesome, thank you!!

You can dynamically keep the sum of the k largest elements in the range you care about with a heap

Store beauty and length in a vector of pairs. Sort by beauty in decreasing order. Now travel from index 0 to n-1. If index < k , then take sum of all lengths till index and multiply it by current beauty and keep max of the answer. Store the lengths in a min-heap or a priority queue with min on top. If index>=k then check where current length is greater than min length in heap or not. If it is greater than min length in heap, delete min from heap and add current length to heap and accordingly change the sum of k top lengths. Then take the required product and keep the max.

sort according to beauty in descending order . Then use priority queue to keep track for k-1 largest length songs .

Sort songs in decreasing beauty, and check for each of them what elements to pick such that length of k or less songs are biggest (and those elements have beauty greater or equal than picked song)

A is so poorly written I spent literally 20 minutes on figuring out what they want from me.. So I it was too late to compete.. Imho almost every educational has some type of really badly written tasks and that you can't understand anything (mostly D)..

I actually thought D was probably one of the most nicely written problems in the set. Their definition of triangulation was a bit different than what I'm used to, but triangulation of a polygon is a very widely used thing that you should know.

I said

mostly D. I was pointing out on previous contests.A: 1hrB: 1hr and 5 min. I need major Improvement. ;)In my opinion, none of the problem statements were badly written, they presented exactly what they wanted to. The matter of fact is that if somebody finds it difficult to understand the task, he marks it as badly written. Believe me, not only is solving a task a skill that comes from practice but understanding a task quickly also comes from practice and experience. So don't blame the statements. Sometimes statements are just complex and not "badly written". Just because you are taking time to analyse them and sort everything out in your head, doesn't mean they are badly written.

Statement of B was confusing, will give more attention to statements in future contests.

Exactly what I was doing the answer would be zero if (s[0]=='>'||s[n-1]=='<') otherwise 1

Now that I saw solutions of other people, Let me explain. for this case :

string : <<>> According to us, answer should be 1 but actually lets say when you delete second character, string becomes <>> then you have to delete first character too. Similary if you delete first charater, string becomes <>>, again you have to delete the first character, so answer must be 2 but according to us, it's 1.

It's my mad during the contest. However I strongly feel that problem statement could have been much much better with proper test cases. I think the intent of statement was to confuse people to write wrong code. Contest should be unrated considering A & B.

You can choose same character any number of times. But that doesn't imply the answer will either be 1 or 0.

Yes, my bad. Sadly I get this after the contest. Will make more test cases in future. Thank you.

don't trust how you understood the problem always read the Note that explain samples ... I did this mistake several times but now I will always read samples note

Somehow I found A much harder than D.

Maybe it's right.

I found all problems harder than D, yeah. D < A < B ~ C

Find solution for

D:1 minuteThink why solution for

Dis so easy:9 minutesExactly.

Well I used argument exchange, so I'll add 2 more steps.

Find solution : 1 minute

Prove solution : 1 minute

Doubt your logic and problem reading skills: 7 minutes

Code and submit : 2 minutes

In problem C I thought that I need to pick adjacent songs only... I was implementing logic similar to KMP and completely wasted 30 minutes lol

weird how the examples show that's clearly not the case?

That shows I was not careful enough

well at least you got

`accepted`

, unlike meDon't compare yourself to other guys, you are just you. Have fun on studying algorithms

Well Bro It is mid night now in East Asia, You are from Georgia Tech so by default I think you are in Atlanta USA, solving problems in mid night sometimes will be pain in ass cause your brain is not functioning well

actually on Betelgeuse, there is no time here

made the same mistake :(

F was very nice.

Is the solution:

My idea ??Try to solve only incremental version, And then use dynamic connectivity trick

Maintain dsu for x and y and then merge them accordingly respectively. Also maintain some links between x and y which are connected. There will be some 4 cases I guess. And update asnwer accordingly.

The easiest way to maintain all components is to create a separate vertex for each x-coordinate and y-coordinate, and treat each point as an edge that connects its x-coordinate with its y-coordinate. Then the answer for each component is the product of the number of x-coordinates and the number of y-coordinates in that component.

How do you erase?

I think this is for incremental version. For dynamic version you have use the idea of dynamic connectivity combined with this.

Just like teja349 has written, I get rid of erase operations by applying divide-and-conquer on time axis (just like in classic Dynamic Connectivity Offline problem).

Just for the sake of curiosity, can problem E be solved if we also ban palindromes of even length?

May I ask why I can see some guys hack them selves by some default inputs? How can this be helpful for them to hack themselves?

If I look for something good in this competition, perhaps that will be that the test case wasn't weak.

Problem E?!!!

I think I understood it incorrectly, can someone please explain it here, It has poor statements.

I think it's the way to fill all -1 in the array so that there isnt any value of i such that a[i] = a[i+2]

We have to calculate the total number of ways to assign values (in range[1, k]) to the -1's in the array, such that no palindrome of odd length is present in the resulting array.

Ok thanks

Can't deduce the Problem B logic Correctly!! Doesn't the problem statement means that every string we have, after some number of operations will either become "<>" or "><"? So the number of removal will either be 1(in case "<>") or 0(for this case"><").

This was the logic I applied. But it seems like I'm missing something. I built this logic because there is no limit to number of operations that i can apply on any given string. So why is this logic wrong?

LINK TO MY SUBMISSIONPlease help!!We have to remove first, then apply operations

If the original string is like "<<>>", then although remove the most out "<" (or ">"), there is another inner "<>". so the correct answer is 2, not 1.

I get this now. After complete period of 2 hours. Thanks Buddy!! Poorly Written Question played with my ratings :(

What do you mean "Poorly written question"?! It literally says

Before applying the operationsin BOLD in the problem text. It couldn't be more clear than that.It triggers me to no end when people don't read the statement carefully and then they say it's unclear. It's not unclear, you just didn't read it properly.

(╯ರ ~ ರ）╯︵ ┻━┻

it wasn't unclear. i analyzed it poorly. gonna read statements more carefully. ಥ╭╮ಥ

you need to check

boldstatement in the problem.Exactly, it helped me to sort out my confusion. Thanks to the author :)

I know we all solved D by some trial and error, guessing, etc... then how can we prove the solution? Any proof?

The edge i..(i+1) is a side of the triangle, in any triangulation. I.e: it would always be an edge. So the best way to connect it to 1 as the third node, and therefore create the minimum possible value.

This assumes separate edges are in separate triangles which need not be the case.

The key idea is that vertices $$$i$$$ and $$$i+1$$$ will always share a triangle.

You are going to have to use all edges in the outer polygon in a triangulation, and for each edge you have to choose one additional label, and using 1 works. So the answer is simply

I don't think this proof is correct. This assumes all edges are part of separate triangles. But if there is a triangle with vertices $$$i$$$, $$$i+1$$$ and $$$i+2$$$, then two consecutive edges are accounted for in the same triangle.

That's true, some of the outer edges can be used together. The triangulation I'm saying is optimal is $$$(1,2,3), (1,3,4), (1,4,5), ..., (1,n-1,n)$$$. While intuitive that this is optimal, what I wrote was not a correct argument.

Actually not

allsolved it by trial and error. I proved it during contest time :)My guesses was, each edge will hide one vertex, so I will hide the biggest one firstly, then the second biggest one, and so on, so they will appear at most two times.

But still the above comments are the proof, IMO.

Actually I tried all possible states xD 51705295

it seems i have proved this solution. Suppose there isn't any diagonal from 1 in optimal triangulation. Then we can move this triangulation in clockwise and decrease the cost. So there is some diagonal from 1 to i. Let's consider triangulation of "1 to i" polygon. If there isn't any diagonal from 1 we can do how described above and so on. So, to minimaze the triangulation cost it needs that all diagonals starts in 1.

This is beautiful.

This comes in somewhat late, but I haven't seen a full proof in the comments yet.

Let the value of a segment be the product of its endpoint. Then clearly the value of a triange is greater than or equal to the value of any of its sides. Moreover, unless one of the vertices is $$$1$$$, the value of the triangle is greater than or equal to the sum of any two sides. This is because $$$abc - ab - ac$$$ factors as $$$a((b - 1)(c - 1) - 1)$$$ which is non-negative unless $$$b$$$ or $$$c$$$ is $$$1$$$.

Now note that every side of the polygon belongs to a triangle in the triangulation. If this triangle contains no other side then its value is at least the value of that side. If it does contain another side then the value of the triangle is greater than or equal to the sum of these two sides, unless one of the vertices is $$$1$$$. By adding up it follows that the sum of the values of all the triangles is at least the sum of the values of the sides, except for the two sides that include vertex $$$1$$$ (because they may be included in a triangle that has another side of the polygon).

Triangulation $$$(1, 2, 3), (1, 3, 4), (1, 4, 5), \dots, (1, n - 1, n)$$$ gives exactly that sum of values, so that's our answer. Also note that the exact labels on the vertices don't matter, all that is needed is that exactly one label is equal to $$$1$$$.

banning mkisic_ from point buff

huh?

You can see his code. It's has a case that cout "YAY YOU HACK ME :D"

Ban Abu_musa_99 too for the same reason. and aihay who hacked themselves with a similar trick.

I couldn't find a reason why problems D had so many AC's. None the less I found that this was a direct answer to the problem. Actually, Problem D was easier than the problem here, but just a bit of modification to the code would do the logic.

Maybe this is the way, Problem D was supposed to be solved, using that DP approach with O(n^3).

That's not really true. We just greatly underestimated contestants. Personally, I thought that majority will stuck in some sort of dp, so after some discussion we decided to leave a place to a straightforward solution. In the end community surprised me in the good way.

PS: linear is not a limit: OEIS/interpolation gives us closed formula and $$$O(1)$$$ solution.

Codeforces Community is Strong. It was a nice contest by the way. Thanks

It’s unfortunate because I would’ve been stuck in a DP type solution if I wasn’t thrown off by the fact that many people were solving it within s few minutes, so I realized it must be some type of greedy solution. If it was a contest with a frozen scoreboard I think a lot fewer people would have solved it.

Actually you can solve problem D in O(1) time complexity.

Let S = 2 * 3 + 3 * 4 + ... + (n-1) * n

3S = 2 * 3 * (4-1) + 3 * 4 * (5-2) + ... + (n-1) * n * (n+1-(n-2))

3S = 2 * 3 * 4-1 * 2 * 3 + 3 * 4 * 5-2 * 3 * 4 + ... + (n-1) * n * (n + 1)-(n-2) * (n-1) * n

3S = (n-1) * n * (n + 1)-6

S = ((n-1) * n * (n + 1)-6) / 3

P/s: I don't know how to use Latex :( So the signs and numbers may be ugly

I agree, It was trivial. I proved this after the contest was over. Always triangulating from vertex 1, works.

Can someone explain how to solve C and the reason behind it?

Okay, the main intuition behind the answer can be said that, if the beauty is fixed, then you can easily find the answer, by taking the sum of all the largest k songs (sorted by length). Using this intuition, we iterate over through all the different beauties present in the array and dynamically maintain the sum of largest k elements, using a priority queue, and updating the sum while traversing. Here is the link to my code — https://codeforces.com/contest/1140/submission/51714832

I think that the following problem resembles this one ?: link

Got it. Thanks!

Hello, Div.2, my old friend

I've come to solve your tasks again...

Can we solve C using DP by iterating on k?

If we can't use DP , can you explain why not?

Probably not since you would have (3*10^5)^2 states

can anyone explain why the output of <>><<< is '0' ? i don't know how to solve this question whats the logic. maybe i misunderstood the problem please explain anyone

For this problem you can always choose the last item to operate on:

After 1 operation: <>><< After 2 operations: <>>< After 3 operations: <>< After 4 operations: << After 5 operations: < Done.

The rightmost < will delete all the values to its left. Hence No need to remove anything and answer is 0

then when will be the answer non zero in which case? so according to you we will always get 0 or 1. but more the 1 values are also there in the output . please explain!!

The answer will be zero if string_var[0] == '>' or string_var[n-1] == '<' else answer will be

min( after_how_much_chars_from_start_appears('>'), after_how_many_chars_from_end_appears('<'))

Like for <<<<>>> Answer will be 3, Read the problem statement again and READ what is in

BOLDcheck this

0ahmad0makki0is fake account ofdotorya. Same solution and also same template.this is also pointed out above, but for some strange reason all the comments are heavily downvotes.

It's not my account, and I think that my account was hacked. (Password leak or something) Now I changed my password, and hope it won't happen again.

May be you are not dotorya and still personating him after hacking his account lol

I can't believe people are upvoting this. Why would dotorya make a fake account and then compete with both of them in the same contest submitting same solutions? I don't think a legendary grandmaster would be that stupid. It must be a hacker's account, should be banned soon.

How to solve G?

I thought the answer could be only "1" or "0",and it took me a while to understand the meaning of

Before applying the operationsin problem B. ;PAny hint for G?

Let's divide all edges into two types: edges of the first type connect vertices having different remainders modulo $$$2$$$, and edges of the second type connect vertices having same remainders.

First of all, let's solve an easier version of the problem: we have to minimize the number of edges of the second type in our path, and minimizing the length of the path has lower priority. Then we can build one tree where each vertex is a union of vertices $$$2x + 1$$$ and $$$2x + 2$$$ in the original graph, and use something like binary lifting or centroid decomposition on this tree. For example, if applying binary lifting, we may maintain an array $$$up[N][LOGN][2][2]$$$, where $$$up[v][k][f1][f2]$$$ is the length of the shortest path that starts in vertex $$$2v + f1$$$ of the original graph, goes $$$2^k$$$ steps up the tree, and ends in the vertex having remainder $$$f2$$$ modulo $$$2$$$ in the original graph.

Okay, but we can't use this approach in the problem from the contest, because sometimes it is better to traverse some additional edges of the second type, so we may choose cheaper edges of the first type. To handle this, let's preprocess all edges of the second type in such a way that cost of each such edge will be equal to the shortest path between its endpoints. This can be done with a Dijkstra-like algorithm.

Thanks for the solution！

How can we made this part "To handle this, let's preprocess all edges of the second type in such a way that cost of each such edge will be equal to the shortest path between its endpoints. This can be done with a Dijkstra-like algorithm." ?

Let's create a graph with $$$n + 1$$$ vertices. $$$i$$$-th vertex will represent an edge between vertices $$$2i - 1$$$ and $$$2i$$$, and in the end, we want the distance from vertex $$$0$$$ to vertex $$$i$$$ be equal to the length of the shortest path between $$$2i - 1$$$ and $$$2i$$$.

How can we find the shortest path between $$$2i - 1$$$ and $$$2i$$$? We either pick the edge between them, or go to some other neighbor of $$$2i - 1$$$ (let it be $$$2j - 1$$$), take the shortest path between $$$2j - 1$$$ and $$$2j$$$, and take the edge from $$$2j$$$ to $$$2i$$$.

So, we need to create the following edges in our graph: $$$n$$$ edges going from $$$0$$$ to $$$i$$$ and having weight equal to $$$w_{2i - 1, 2i}$$$ in the original graph, and for every pair of odd neighbors $$$2i - 1$$$ and $$$2j - 1$$$ an edge going from $$$i$$$ to $$$j$$$ and having weight $$$w_{2i - 1, 2j - 1} + w_{2i, 2j}$$$.

All that's left is to run Dijkstra from vertex $$$0$$$ in this graph, and that's how we get preprocessed weights.

if The

CF-Predictorbroken ? ( Chrome Browser)=>Use this Website=>About the IssueSecond problem ruined the whole contest for me. The problem statement must be properly written.

Same with the first one...

UPD: But the problems were great! :)

ExampleandNoteare very useful to understand the problems.Please update ratings :3

Waiting for hours I've been. Painful it is.

What's the approach for F?

I have the following observation but I didn't solved it.

Consider vertex $$$(x, y)$$$ as an edge connecting $$$a_x$$$ and $$$b_y$$$. Then the answer is the sum of (number of $$$a_i$$$ $$$\times$$$ number of $$$b_j$$$) in each connected component. If we can maintain the connectivity supporting deletion efficiently, the original problem can be solved.

Edit: I upsolved it by doing divide-and-conquer on time axis. Distribute the lifetime of each vertex to $$$\text{O}(\log n)$$$ time intervals. Then traverse the segment-tree-like structure with disjoint set union operations and restore operations.

Can you give more explanations? It's hard for me to understand your code))

Firstly, suppose a vertex exists during $$$[2, 9]$$$. Then you can split the interval similarly to range query in segment tree, i.e. $$$[2, 9] \rightarrow [2, 3] + [4, 7] + [8, 9]$$$ (it may vary due to different implenmentations of segment tree/divide-and-conquer).

Call the following function recursively.

It is easy to see that:

each answer is printed in correct order

all vertices have the correct state (existing or not) when each answer is printed

there are almost $$$\text{O}(n \log n)$$$ operations of adding vertices

consequently, there are almost $$$\text{O}(n \log n)$$$ undo operations

It's amazing. I have understood the algorithm and the code!))

please update the ratings.

PLEASE UPDATE THE RATINGS!

Can someone explain their approach in solving C, I dont know why I got hold of a very wrong track in solving it, and finally it became a mess.

Let $$$x = [(\text{beauty}_{k_{1}}, \text{length}_{k_{1}}), (\text{beauty}_{k_{2}}, \text{length}_{k_{2}}), ...]$$$ This array is sorted by beauty.

Then if you pick $$$(\text{beauty}_{k_{i}}, \text{length}_{k_{i}})$$$ as minimum beautiful song, then you must pick up to $$$k-1$$$ songs from $$$[x_{i+1}, x_{i+2}, ..., x_{n}]$$$. You can use

`std::multiset`

or something to pick $$$k-1$$$ most longest songs.When selecting a song, the total length and minimum beauty matters. So let's sort the songs by decreasing beauty, and process them like that. To make sure we have maximum length, we keep a multiset of the k-1 longest songs processed so far. Let's process song i. We do ans = max(ans, max_total_length*beauty[i]. If the multiset has size less than k-1, add this song's length to it and increase max_total_length. Otherwise, if this song's length is greater than the shortest length in the multiset, remove that length, subtract it from max_total_length, and add the current like previously.

I have a doubt , why don't we take k longest processed so far and do the all think you explained?

Please update the rating.

I think system testing is still due

Please update the rating.

Please update the rating!

PLEASE UPDATE THE RATINGS !!!

Have refeshed my page hundreds of times.

Why not charge in the rating?It is really disappointing that some 2100+ rated players make fake accounts before the contest in order to participate officially. For example take a look at this account: bigbigbigbigcat111 This account was made just before the contest and ranked 30. I think this account was made by bigbigbigcat111 who is unable to participate officially because his rating is above 2100. Making fake accounts and stealing rating from other participants is unacceptable. Please remove him from the list of official participants. Upd: I noticed that he has account: bigbigcat111 and bigcat111 as well. All accounts are from China and they have similar rating.

What happened to cat111?

Please, someone should steal

bigbigbigbigbigbigbigbigbigcat111to stop this unethical behavior!Please do not refresh the ratings... I'm sad enough...

...Why hasn't it been rated yet?I've been waiting for a long time.Just be a little faster,OK?Thanks!

Probably there's alot of cheaters in this round that's why the system hasn't started testing.

When will the editorial come out?

Probably after the System Testing gets completed I guess

What about the System Testing, when it's going to start?

It has started.

Why is system testing is so late?

System Testing ended.

problem c was very nice,thanks for the problem author!!!

When the editorial will be published?

Hello!Can anybody expalin why my sol for problem C fails?Seems like it must work.Thanks! https://codeforces.com/contest/1140/submission/51763978

If you use tmp.erase(x), it will remove all the elements with value "x" from the multiset.

You have to remove this "if", because of the following case:

Case:

https://codeforces.com/contest/1140/submission/51766475

thank you very much!

Auto comment: topic has been updated by PikMike (previous revision, new revision, compare).How can we solve D? I think I should solve prefix sum... T^T I will do not go to sleep when I solve D before!!