Codeforces Round #774 (Div. 2)

3 years ago, # ^ |

-94

Why doesn't Ukrainian contestants boycott this Russian website?

→ Reply

qwexd

3 years ago, # ^ |

+14

use grammarly

→ Reply

izuku_midoriya

3 years ago, # ^ |

how did you reach candidate master without common sense?

→ Reply

AsishKumar

3 years ago, # ^ |

← Rev. 2 →

down! Charon here

→ Reply

Maango16

3 years ago, # |

← Rev. 2 →

Note The Unusual Timings (1 Hour Late)

→ Reply

swoon

3 years ago, # |

gl ^ hf

→ Reply

__declspec

3 years ago, # ^ |

+25

good luck ^ have fun = 0

→ Reply

valeriu

3 years ago, # ^ |

This was such a hint and everybody ignored it smh

→ Reply

Amir_

3 years ago, # |

best of luck for all♥️

→ Reply

sdyakonov

3 years ago, # |

← Rev. 2 →

-15

Good luck and positive rating!!!

→ Reply

GaGeV

3 years ago, # ^ |

-18

why are you disliking him?

→ Reply

sdyakonov

3 years ago, # ^ |

Mike Top. I wanted to make a joke

→ Reply

Olympia

3 years ago, # |

This will probably be my first contest in over a month...I'm awaiting -100 delta :P

→ Reply

SirPh

3 years ago, # ^ |

← Rev. 3 →

same... I'm awaiting losing my green after my first contest :P gl

that wasn't true gonna get +20 :yayy:

→ Reply

notwatermango

3 years ago, # ^ |

← Rev. 2 →

let's go! i'm having my series of defeat too

i overslept so i didn't participate

→ Reply

gagan_sharma

3 years ago, # |

+10

Good luck everyone. Maybe this round will help improve people's mood at codeforces.

→ Reply

rishisgsits_22

3 years ago, # ^ |

-20

hello gagan plz guide me as i am stuck in newbie-pupil phase from a long long time and i don't see improvement also provide tips for dp if any. thnx in advance.

→ Reply

gagan_sharma

3 years ago, # ^ |

← Rev. 2 →

The key to improvement is indeed just good-old practice. Now, everyone should choose a practice regime which suits them well. I just like you were shuffling pupil and newbie for 1 year. This year I started practicing more with a focus on solving questions rather than learning new algorithms. You only need a handful of algorithms like dfs-bfs, Dijkstra, dsu, segment tree to reach the expert but dp practice is very important. I practiced dp from problemset with filter-tag and started solving from 1000 rated questions and gradually increased rated questions. When I felt like I can do 1600 rated dp questions pretty confidently then I started randomly doing questions with gradually increasing difficulty from 1000 to 1600.
Make sure you actually solve the questions and don't see the editorial until you solve it. If it seems impossible then try to read other people's code before checking out editorials. Do cp with your friends, discuss approaches and try to outrank your peers :p. This is what has worked for me till now, hope this helps you. Good luck!

→ Reply

3 years ago, # ^ |

I think DP is the most important thing. segment tree is not too important at pupil stage.

→ Reply

gagan_sharma

3 years ago, # ^ |

I was talking about the concepts which can land you to expert level.

→ Reply

Hd7

3 years ago, # ^ |

expert level doesn't need segment tree to be fair, binary search, greedy, simple DP, DFS/BFS and a little knowledge of graph.

→ Reply

Sakura_mw

3 years ago, # |

Maybe we have the opportunity to just focus on this round itself!

→ Reply

Aromyase

3 years ago, # |

Look at the quality of the questions. My hands are too slow.Orz

→ Reply

taogehtsilalos

3 years ago, # |

Before the usual time was 16:35, then 15:35, then 11:35 ... What is the actual usual time ?

→ Reply

Mohd_Arbaz

3 years ago, # ^ |

9:05 PM in IST, really good for indian contestents, we don't have to shift our dinner this time.

→ Reply

super_sage

3 years ago, # ^ |

Tatakae

→ Reply

DimmyT

3 years ago, # ^ |

TATAKAE! TATAKAE!!

→ Reply

jakubd

3 years ago, # ^ |

15:35 winter time, 16:35 summer time

→ Reply

Murot_06

3 years ago, # |

ooo thanks

→ Reply

AHMADUL

3 years ago, # |

← Rev. 2 →

Are we going to fight against "Robot" in this round?

→ Reply

vaibhav.goyal

3 years ago, # |

May C problem be solved by less than 2000 people. I want it to be an interesting one.

→ Reply

JEE_MAIN_LOOSER

3 years ago, # ^ |

Bruh This is every Cyan dream . That it has less than 2000 solves and we are among them . But sadly cheaters will never make it happen ;_;

→ Reply

TimDee

3 years ago, # ^ |

I would say and pupil too — last 3 contests happens to me top 1000 after A,B (top 100 last round), and then somehow stuck on C, and -∆, -∆, -∆

→ Reply

turkoarias

3 years ago, # |

+59

First argentinian contest! Congratulations MateoCV

→ Reply

rohitsatpute480

3 years ago, # |

Contest after long time.Eager to participate

→ Reply

bernborgess

3 years ago, # |

Good luck everyone! Have a good contest!

→ Reply

JEE_MAIN_LOOSER

3 years ago, # ^ |

You too my friend !

→ Reply

Daniel19Frenzy

3 years ago, # |

Argentinian round? Aguante el Diego, Talleres y FaMaF!

→ Reply

mily

3 years ago, # |

Good luck!

→ Reply

F99_ll_H21

3 years ago, # |

Unfriendly time for Chinese contestants.

→ Reply

3 years ago, # ^ |

you can't make everyone participate at a good time, right?

→ Reply

F99_ll_H21

3 years ago, # ^ |

Of course.Good luck to all contestants around the world.

→ Reply

jakubd

3 years ago, # ^ |

← Rev. 2 →

-14

dead meme

→ Reply

StArChAn

3 years ago, # |

+20

Dear contest, How do you feel today? Yet another person commented on your announcement page rn (me). Since I said hi to you, please give +ve this time to me. yk the usual. thanks in adv

Regards, that random user asking for +ve in rating

→ Reply

BelalMoawad

3 years ago, # |

← Rev. 2 →

Very cool scoring distribution 1000 — 750 = 1250 — 1000 , 3000 — 2500 = 2500 — 2000

→ Reply

happy_new_year_2022

3 years ago, # ^ |

Dam bro, did you just invented arithmetic progression.

→ Reply

b4nanacream

3 years ago, # |

there is no as a tester comment.

→ Reply

AgentNinja

3 years ago, # |

SpeeeeedForcesssss

→ Reply

3 years ago, # ^ |

why? the number of people who solved C isn't high or too low

→ Reply

early-morning-dreams

3 years ago, # |

antonforces

→ Reply

vrintle

3 years ago, # |

+57

→ Reply

Aoi_Hill

3 years ago, # ^ |

+11

D seemed near yet was so far.

→ Reply

3 years ago, # ^ |

Same with E :)

→ Reply

3 years ago, # |

I haven't coded recursive code in 2 years lol good to remember it with problem C

→ Reply

bluescorp

3 years ago, # ^ |

I did it without recursion

→ Reply

Amrit_4343

3 years ago, # ^ |

How?Plz share your approach..

→ Reply

bluescorp

3 years ago, # ^ |

← Rev. 3 →

Store the factorials numbers <=1e12 in an array (I think the count of them is 13 or 14)

Just iterate on all subsets of them (by iterating from 0 to 2¹⁴) and then subtract the factorials number from n (let's call the resulting number as temp). After each subset, count the number of set bits in temp. This will be the number of powers of 2 required. Do this for all and keep track of min so far. Submission https://codeforces.net/contest/1646/submission/148374185

→ Reply

sohelH

3 years ago, # ^ |

you mean factorials?

→ Reply

amitroy7781

3 years ago, # ^ |

Solved it using bitmask dp on factorials <= 1e12 and pop count

Submission: https://codeforces.net/contest/1646/submission/148339954

→ Reply

I_LOVE_UNDEFINED

3 years ago, # ^ |

I don't think there is any DP in your solution :p.

→ Reply

amitroy7781

3 years ago, # ^ |

sorry used bitmasks to generate all possible subsets of size 15 :)

→ Reply

amitroy7781

3 years ago, # ^ |

Why did the setters confuse with -1 where ans is always possible?

→ Reply

JEE_MAIN_LOOSER

3 years ago, # ^ |

Just to check that we are smart enough to think that . As its quite obvious that every no can be represented in binary form. So there no chance for answer to be -1.. But some ppl i saw do handle that case ;_;

→ Reply

3 years ago, # ^ |

I think it's still the same complexity recursion can be done in a different approach of course

→ Reply

3 years ago, # ^ |

I did with a mask-compressed DP...

→ Reply

hardikchopra

3 years ago, # |

It took me 2WA's and 50 minutes to figure out that for long long integers __builtin_popcount doesn't work and we need to use __builtin_popcountll ... :-/

→ Reply

3 years ago, # ^ |

I realized it just before click submit button...

→ Reply

hardikchopra

3 years ago, # ^ |

I am just happy that at least I figured it out before the contest ended :)

→ Reply

Aoi_Hill

3 years ago, # ^ |

← Rev. 2 →

-7

You should keep such things as #define (macro), easier to remember

→ Reply

memoryallocator

3 years ago, # ^ |

← Rev. 3 →

I suggest using std::popcount(). The only issue with this function in terms of competitive programming is that it requires x to be of unsigned type, so we have to cast explicitly if x is long long.

int popcnt = popcount(uintmax_t(x));

Of course, the simple macro solves this problem: #define popcnt(x) popcount(uintmax_t(x)).

→ Reply

Zhora_.004

3 years ago, # |

+17

5th test case of problem D...

→ Reply

ADO

3 years ago, # |

it was so fun to write 15 nested for loop in problem C(p.s. I know recursive solution would also work)

→ Reply

b4nanacream

3 years ago, # ^ |

that means you don't know this

→ Reply

ADO

3 years ago, # ^ |

I mentioned fun, actually you don't know what I know so please read comments carefully next time

→ Reply

3 years ago, # |

← Rev. 4 →

+67

Imagine wasting 20 mins debugging E because you "simplified" $$$\frac{lcm(x_1, x_2, \ldots, x_k)}{x_1}$$$ to $$$lcm(x_2, \ldots, x_k)$$$ without thinking properly T_T.

Also I can't exactly describe it, but CDE feel like they are somehow standard yet fairly interesting at the same time. Maybe its because they have extremely simple and natural formulations which don't feel like they've been forced to match an idea?

→ Reply

puffinmuffin

3 years ago, # |

A nice contest for me!! I loved the problem C. Would love to share my experience while solving problem C,

Initially after reading the problem , I immediately thought of solving it using subset generation through bitmasks , because i thought the size of factors + powers of 2 would be around 30 , but turned out it was ~52. So i drifted a bit & thought of solving it using DP(using coin change approach) but that is also not feasible.

Solution :

Spoiler

→ Reply

3 years ago, # ^ |

you only need to bitmask factors, then greedy choose powers of 2 for the rest.

→ Reply

puffinmuffin

3 years ago, # ^ |

Yup, thats what i did. I don't think you need to greedy choose powers of 2 for the rest , because the remaining number's binary representation containing the set bits already is the least number of powers of 2 required to represent it.

→ Reply

codersattu

3 years ago, # |

Logic for C?

→ Reply

3 years ago, # ^ |

Note that n can allways be constructed with number of its bits, since these are powers of 2.

So construct all possible sums of factorials (that are at most 2^14 since there are only 14 factorials in range), subtract each one from n and count the bits of the difference.

Smallest number of set bits in difference plus bits in factorial sum is ans.

→ Reply

pooty

3 years ago, # |

How to do F?

→ Reply

Rudro25

3 years ago, # |

What d hell is test case 5 of D -_-

→ Reply

3 years ago, # ^ |

+11

Either a large graph or something that kills the incorrect bipartite idea I guess.

Try the following graph if you don't get why bipartite is incorrect:

Input Graph

Answer

8 10
1 1 1 1 1 1 1 1 1 1

→ Reply

Rudro25

3 years ago, # ^ |

Thanks sir. got my mistake <3

→ Reply

Restricted

3 years ago, # ^ |

There are cases where it is optimal to not take two neighboring nodes. Splitting the tree to a bipartite graph and then taking one of the two parts is not always optimal. The idea is similar to the well-known DP problem "Maximum sum over array such that no two elements are adjacent" but implemented on a tree.

→ Reply

ding_don_dino

3 years ago, # ^ |

+24

2,3-dimethylbutane

→ Reply

fonmagnus

3 years ago, # |

My Idea for D :

Take all leaves to be good vertices
Now all parents of the leaves (call it buds) are assigned with weight = 1
Now the graph is split into separated sections with "buds" as boundaries for each section
For each section, I solve them independently by coloring it "bipartitely". In each section, I find which gave me the best arrangement whether the color "black" / "white" on this section are good vertexes and bad vertices for its counter-color.

I got WA on pretest 6. Not sure if the idea is wrong or my implementation My submission : 148381092 Any thoughts?

→ Reply

kwangg

3 years ago, # ^ |

I tried the same thing (but failed pretest 5) :(

→ Reply

shivakumar.nannan

3 years ago, # ^ |

it is true that a good vertex(blue) must be surrounded by bad vertices(black) but it is not necessary that bad vertices be adjacent to good vertices. so if u two color the tree, u wouldn't take into consideration cases where two adjacent nodes are bad

→ Reply

45muillaht

3 years ago, # ^ |

Pretty sure it's finding the max independent set.

→ Reply

Eren44

3 years ago, # ^ |

← Rev. 2 →

I did it in sort of dp style. Rooted the tree at 1.

dp[i][0] = <C,S> means ith node is not good and fill its children however you want such that max good vertices below my subtree is C with sum of edges as S.

dp[i][1] = <C,S> means ith node is good and fill its children however you want such that max good vertices below my subtree is C with sum of edges as S.

For transitions dp[i][1] depends on dp[c][0] where c is children of i since you are good, children can't be good.

dp[i][0] picks dp[c][0] or dp[c][1] according to which is best.

→ Reply

shivakumar.nannan

3 years ago, # ^ |

did it work ? i tried the same thing but kept getting WA in pretest 5 or 6

→ Reply

45muillaht

3 years ago, # ^ |

You also need to keep track of the cost in case dp[i][0]==dp[i][1].

→ Reply

Eren44

3 years ago, # ^ |

Yeah the reason for storing S in the dp is to handle the case of dp[c][0] == dp[c][1]

→ Reply

3 years ago, # ^ |

← Rev. 3 →

How does the problem change after removing all the leaves?

If there exists some tree for which bipartite doesn't work, I can just create a new tree by attaching a single node to each leaf and your soln won't work for it.

And try the following graph if you don't get why bipartite is incorrect:

Input Graph

Answer

8 10
1 1 1 1 1 1 1 1 1 1

A Correct Solution

→ Reply

shivakumar.nannan

3 years ago, # ^ |

i defined dp[i][0] = max good vertices if vertex i is good and dp[i][1] for bad. and dp[i][0] = 1 + all dp[ne][1] in neighbours and dp[i][1] = max(dp[ne][0] and dp[ne][1]) for all neighbours. Is is correct ? i kept getting WA in either pretest 5 or 6

→ Reply

Eren44

3 years ago, # ^ |

I guess in dp[i][1] case if dp[ne][0] and dp[ne][1] are equal then how are you ensuring which one to pick to get minimum sum.

→ Reply

3 years ago, # ^ |

Yup, this is likely the issue.

One way to solve this is to store the dp value as a pair $$$<\text{max number of nodes}, -(\text{min sum of degree})>$$$, then taking max works as expected.

→ Reply

fonmagnus

3 years ago, # ^ |

Oh I see. I get it. Thanks!

→ Reply

fonmagnus

3 years ago, # ^ |

Wait, in your explanation should it be $$$dp_{x, 0} = max(dp_{v,0} , dp_{v,1})$$$ ?

→ Reply

Girus

3 years ago, # |

← Rev. 2 →

Could someone share their approach for B ?

I tried solving it by sorting the array (minimize the Blue sum) and using prefix sum, but failed. I think this is because my approach colored all the numbers.

→ Reply

3 years ago, # ^ |

Use sum of two smallest numbers as small sum, and biggest number as big sum.

If both sums are ok ans=YES else add next smallest number to small sum, next biggest number to big sum. Repeat.

→ Reply

_sh0x

3 years ago, # ^ |

Well, I also sorted it first, and then kept on calculating Blue Sum from the start and Red Sum from the end of the array. Also, to maintain the count(Blue) > count(Red), I initialised the starting index to be 1(0-based indexing) while initialising Blue Sum to be equal to the first array element.

→ Reply

Yaria

3 years ago, # |

← Rev. 3 →

submitted C in the last minute

took more than 1 hour to realize brute-forcing all subsets of factorial values is feasible

but brute-forcing all subsets of 2-power values is infeasible which gave me a wrong intuition that the factorial counterpart is also infeasible

I hope in the future I can be more familiar with the thresholds of feasibility.

→ Reply

snehan1234567

3 years ago, # ^ |

Why we Can't use one factorial two times?

→ Reply

Yaria

3 years ago, # ^ |

because we need distinct numbers

→ Reply

snehan1234567

3 years ago, # ^ |

Got it now. I Read the question again. Sucks when missed important point.

→ Reply

3 years ago, # ^ |

That would make it way harder so be glad about it

→ Reply

3 years ago, # ^ |

the problem description said distinct.

→ Reply

m_vid

3 years ago, # ^ |

Even if it was feasible to brute force all powers of 2, how would you find the needed amount of factorials?

→ Reply

Yaria

3 years ago, # ^ |

I don't know how to find the needed factorials except for brute-force, so I should have taken that approach initially. Unfortunately the infeasibility of brute-forcing 2-powers gave me the wrong intuition that brute-forcing factorials is also infeasible.

→ Reply

snehan1234567

3 years ago, # ^ |

No Idea. That is why I was not able to solve C. Was expecting there is some other trick to solve the problem which I was not able to think.

→ Reply

3 years ago, # ^ |

There are not much factorials in range, it is only 14 numbers. So you can create only 2^14 sums from these numbers. Just try all of them.

→ Reply

agrmayk

3 years ago, # ^ |

Went through the same trajectory, except I took even more time to realise this and couldn't submit my code in time :( It also somehow escaped by mind that combinations of powers of 2 literally define every single number in exactly one possible way.

→ Reply

E404_Not_Found

3 years ago, # |

Thanks for the contest, I don't know why I did so dumb on A and B, But after all I really liked the problems.

→ Reply

quiet_one

3 years ago, # |

In c, finding all the numbers powerful in a set then take a number which are just less than n and subtracting them from n is not feasible why?

→ Reply

barun511

3 years ago, # ^ |

← Rev. 2 →

For 184 the correct answer is 120 + 64 but your approach would yield 128 + 32 (edited from 56) + 24.

→ Reply

TimDee

3 years ago, # ^ |

How did you get 56 here....

→ Reply

barun511

3 years ago, # ^ |

184 — 128 = 56. The point is that you can't pick the max number <= n.

→ Reply

TimDee

3 years ago, # ^ |

But 56 won't be in set, so it would look like 184 → 128 + 32 + 24 (wrong approach)

→ Reply

barun511

3 years ago, # ^ |

Ah sorry, I meant 128 + 32. Thanks :)

→ Reply

AdityaG

3 years ago, # |

what is pretest 6 of D? ಥ_ಥ

→ Reply

kanishk779

3 years ago, # ^ |

If you were doing BFS solution from leaves, It would fail. Optimal solution will be reached by dynamic programming.

→ Reply

pcajourney

20 months ago, # ^ |

Can you please explain why bfs from leafs doesn't work ?? Any test-case ???

→ Reply

adi1998

3 years ago, # |

← Rev. 2 →

+14

I think D is just Maximum Independent set with least total degree which can be solved by dp. I realised this, but could not reconstruct the solution from the dp before the contest ended.

→ Reply

chika10

3 years ago, # |

How to solve E? I was looking for numbers with the same prime factors

→ Reply

3 years ago, # ^ |

Take 2, 4, 8, ... as example. In the row starting with 2, you get 2^1, 2^2, ... In the row starting with 4, you get 2^2, 2^4, ..., so what in the cells are these powers of 2: 1 ~ M, 2 * 1 ~ 2 * M, 3 * 1 ~ 3 * M, ..., depending on how many powers of 2 are less than or equal to N. Then, what you need to do is calculate how many distinct number is in this list. Sum up all the length of lists of a number that is not power of another one. Don't forget 1.

→ Reply

Alireza

3 years ago, # |

Any counterexample on D solving with 2-coloring?

→ Reply

physics0523

3 years ago, # ^ |

→ Reply

mohamedeltair

3 years ago, # ^ |

1 2

1 3

1 4

4 5

4 6

The output should be:

4 6

1 1 1 1 1 1

The idea is that although there can be no 2 adjacent good nodes, this does not mean that there can be no 2 adjacent bad nodes. Unfortunately I realized that a short time just before the contest ended.

→ Reply

3 years ago, # ^ |

Input Graph

Answer

8 10
1 1 1 1 1 1 1 1 1 1

Visualization

→ Reply

Abanoub_Asaad

3 years ago, # |

← Rev. 2 →

Hey,
In the first problem, why this works cout << s/(n*n) << '\n';
but using cout << (int)(s/pow(n,2)) << '\n'; gives wrong answer in test 2?

→ Reply