Please subscribe to the official Codeforces channel in Telegram via the link: ×

Recent actions

Yes it is

Created or updated the text


Created or updated the text

I don't even understand the problem. Could you help me?

Here is a clean implementation of this approach.

Created or updated the text

same problem

for the same code once gave wa on test case 8 and next time gave runtime error on test case 1? whats the problem with the online judge?

Someone please help me with problem C. I couldn't understand the problem.

I tried to solve problem C with Binary Search but failed.. whats wrong with my algorithm? thx :)

here is my submission 46606310

LooOoLOoOoOooOoLoooOooOoOL. LMAO. ROFL. Might as well don't post your solution here if you are going to post code without any explanation.

python 6-lines AC solution to problem B

_, string = raw_input(), raw_input()
start, end, res = 0, 0, 0
for char in string:
    if char == "G": start += 1
    else: end, start = start, 0
    res = max(res, start + end + 1)
print min(res, string.count("G")) for more

Educational system is a modification to icpc system, not the cf one. Over there all the submissions after AC don't count, the same is here.

Auto comment: topic has been updated by PikMike (previous revision, new revision, compare).

That's odd though. On "normal" codeforces rounds, only the LAST submission to pass the pretests in judged on systests and that one submission will matter towards your score.

How come it's different for "educational" rounds?

(I will use N to denote size of a trie)

Don't know about intended; I did following DP: [subtree that we have to solve (N)][Depth of the first taken vertex on path from subtree root to the overall root(N)][how many vertices do we want to take in this subtree(K)]. So I have N * N * K states, and in order to solve the state I'm running knapsack over all sons of the vertex. Adding up those knapsacks for all vertices, we'll get N * K * K over whole tree (each vertex is a son of exactly one other vertex, and it will be considered in K * K possible transitions in knapsack), plus we have to run it N * K times for all possible combinations on two other dimensions of DP, which leads to the bound of N2 * K3 with a really low constant.

When will tutorial be available?

I see, that makes a lot of sense!

hi there i have a problem. recently, i got a message that said your round is going to be skipped. bu at the end, it said that if you think there is a problem, you can send a message here. i was getting +98 from this contest, so i think there was no reason to skip this round by my self. thanks.

Intended solution for F? I can only bound my solution with , where S is the sum of lengths of all strings (that also may be slightly wrong, to the better or worse), but the constant is incredibly low and I can't come up with a worstcase. It passed with 233ms or so.

I'm still sort of confused about the Educational rounds. Will there be a system test phase after the open hacking phase? Or the contest just finishes once the open hacking phase finishes?

In Educational rounds the systest consists of the hack inputs used during the open hacking phase. If the input needed to hack the code is very specific, it's possible that it passes the other hacking inputs and thus the systests.

Shit just got real...System testing is just lit this time...

Would that really matter though? There will be system test to eliminate those wrong answers if nobody hacked them anyway right?

Ok. Understand it. Thank you!

Yes, you will.

That's sounds reasonable. Moreover, if the first one failed the systests while the second one passed, will I get penalty for the first one?

finally I still can't hack this 1965ms O(n^2) 46324307

50000*50000 just cost 2s.. amazing...

but I hacked another solution with java : 46336111

but in fact this solution's time cost is about 2400ms

so this tells us drop java and pick c++ now

just joking XD

No one cares.

The earliest submitted solution to get AC on systests will be considered.

Ohh, sorry.

Thank :3

dp[x][j — 1] = ind; --> dp[x][j — 1] = lower_bound(g[x].begin(), g[x].end(), ind) — g[x].begin(); FeelsBadMan

I can help you with that :P. In fact, the book "Algorithm Design" by Jon Kleinberg, Eva Tardos can be truly revealing about that.

Go to page 396(Chapter 7 Network Flow). This problem is known in the literature as "Project Selection". Enjoy.

PD: The book i pointed to you is amazing, there are other non straight forward applications. My suggestion: Read other chapters as well.

Think of it like this. Initially we have taken all "edges" to form our subgraph. If we cut an edge connecting source to "edge" it means we don't take the "edge" in our subgraph, similarly if we cut an edge from vertex to sink we don't take the vertex in our subgraph. If an edge is present in our subgraph then the 2 vertices it connects must be in our subgraph too. If an edge is not present in our subgraph then atleast 1 of the 2 vertices it connects must be out of our subgraph. In this network any cut will fullfill these rules(try to see why this is true) and to maximize the answer we just need to find the min-cut.

Thank You

In my defence, however weak it might sound, I primarily code in Python. Which still is not a reason to miss semi-colon.

r always must be < g[a[i]].size(),

because g[a[i]].push_back(i) and r = lower_bound


r always < g[a[i]].size()

why UB???

and...??? why is it wrong???

r = g[a[i]][r]; //UB

oh ;-; thanks!

I think the function is not strictly increasing

Can someone elaborate why this works and what's the intuition behind it?

For problem C I wrote code in c++ in ideone it ran for sample test cases well.But on submitting it is giving runtime error with "Exit Code is -1073741819" for the very first test case I don't know what the exit code means? Can anyone explain why and when that exit code occurs?


Why does Binarysearch WA for C?

Just check your IF statement carefully. you put a ';' just after the if (i < matrix[j].size()) statement which caused the wrong answer.

Hello! idk if this is intended but I found out I could submit for practice before contest ended, this seemed really cool however when I get wrong answer I can see the test data!!! which means I can hack any code I want that gives Wrong Answer, is this the intention for the educational CF?

Another HackerForces contest!!!

Many thanks~ However, I think I should go to bed now because it is already 1: 30 in China now...


You can indirectly get an extra bonus: if you hack someone that is higher than you in standings and he falls below you, you get a better position in final standings and a bit more rank.

Can anyone tell me why this code has been blowing up with undefined behaviour on Codefroces servers ?. It works fine on my machine and . My Submission



This submission 46340090 Wrong answer on test 11,

but on any other compiler 11 test — ok (and on my local machine).

I do not understand why and I think that the problem is in the cf compiler,

help pls.

//11 test//


18 1

3 3 1 3 1 1 3 4 2 2 2 1 2 1 4 1 4 3



And what if I hack successfully? Will I get a extra bonus or only the one who is hacked will lose his "accepted"?



Maybe this will help

It says that "Diagnostics~" issue might be due to undefined behavior.

Make sure that you never try to access index -1 of an array or something like that.

May God save me.

First, if sum(a)<2*(n-1) then you can't build a connected graph at all. Otherwise you need to build a tree that is maximally close to just a path. My approach was like this: first, build a path from vertices with degree 2+, then add a vertex with degree 1 (if there are any) to the first and last vertices of that path. Add other leaves anywhere respecting degree constraints.

how to solve C??

Fix the two vertices with lowest a as the ends of the diameter. Put all vertices with a >= 2 between them. Then check if you can put the remaining vertices (they have degree 1) on them.


could anyone please tell me what is this hell -> Diagnostics detected issues [cpp.clang++-diagnose]? only due to this i could not submit my B problem. although my code is working fine on my system.

Will I get a penalty if I hack unsuccessfully?

How to solve D?

If the first passes it will be considered. Else the second

The last submission will be judged

yeah tricky with false test explaining

Make a flow network and represent each edge as a vertex. Connect each "edge" to the source with capcacity the weight of the edge. Connect every vertex V of the graph to the sink with capacity a[V]. Finally connect each "edge" with the 2 vertices it connects with capacity Infinity. The answer is sum of weights of all edges — min cut in this network.

Is Dinic's algorithm enough for Problem G? It seems that the complexity is something like O(n3).

By the way, I got accepted with Dinic's after I got accepted with Ford-Fulkerson. Would someone tell me which submission the system will use to judge?

Make a flow network and represent each edge as a vertex. Connect each "edge" to the source with capcacity the weight of the edge. Connect every vertex V of the graph to the sink with capacity a[V]. Finally connect each "edge" with the 2 vertices it connects with capacity Infinity. The answer is sum of weights of all edges — min cut in this network.

Really good contest, tricky A though :D

Can anyone share their approach for G? I saw G started getting accepted solutions much before E and F started getting accepted.

that's not fair I actually wasted 15 min on A because i cant understand the second test case until the Global announcement.... maybe I would have solved C I'm just putting final touches on my solution

halyavin is here.

How to solve G??

challenging problems are today.

Auto comment: topic has been updated by PikMike (previous revision, new revision, compare).

Clashes with Carlsen vs Caruana tie-break

"Is it rated?" Life Hack: Read the title!


Yoo!! Another contest by Pik Mike How To lomk Him ? we Didint have Div3 For Long Time And Is It Rated ? that isn’t something to be happy about

sorry xD

Everytime I said "Hope we all have a good rank and I want to become expert"then got a bad rank,so I say hope everyone enjoy this contest this time :D

We haven't seen div3 for long time.

I apologise, I missed that too. Good luck for the contest!

Oh, I missed that. Then it's a long time before Mail.Ru round 3.

Mike was problemsetter at Mail.Ru round 3, which was on Sunday.

Idk if 3 days are so long

(int) Round 55 Div. 2 = 28 November :D


Can something be done please to avoid clashing with the SRM?

except for div.3, yeah! :)

It's been so long since MikeMirzayanov appeared as a problem setter(except div. 3) :)

Yoo!! Another comment by HARRYPOTTER0

Yoo!! Another contest by PikMike

Created or updated the text
Created or updated the text
Created or updated the text
Created or updated the text
Created or updated the text