awoo's blog

By awoo, history, 4 years ago, translation,

1366A - Shovels and Swords

Idea: Roms

Tutorial
Solution (Roms)

1366B - Shuffle

Idea: Roms

Tutorial
Solution (Roms)

1366C - Palindromic Paths

Idea: BledDest

Tutorial
Solution (BledDest)

1366D - Two Divisors

Tutorial

1366E - Two Arrays

Idea: Roms

Tutorial
Solution (Roms)

1366F - Jog Around The Graph

Idea: Neon

Tutorial
Solution (pikmike)

1366G - Construct the String

Idea: Neon

Tutorial
Solution (Ne0n25)
• +113

| Write comment?
 » 4 years ago, # | ← Rev. 2 →   +2 I realized in hacking phase that for E I accidentally read in $a$ and $b$ both with lengths $n$ (see my 83457987). I thought I would segfault in system testing but I was still accepted. Can someone explain from first principles why this is? I guess it's ok to cin out of bounds as long as you don't use the memory afterwards?
•  » » 4 years ago, # ^ | ← Rev. 2 →   0 getting tle on case 46 for E on this......Can anyone tell why? upd: solved
 » 4 years ago, # |   +12 Hey awoo there is some problem with font of code of f because of comma in mod
•  » » 4 years ago, # ^ |   +8 and e too
•  » » » 4 years ago, # ^ |   +12 Fixed both, thanks
 » 4 years ago, # |   +5 Can some one please explain me how to do the Linear Integer programming, here in Problem A, they are checking just the corner points of the graph drawn & deciding on the answer,and yes it is quite true when x1,x2 are real numbers, but here the constraint is x1,x2 are integers>=0, so how to approach this problem. please help me with the proof.
•  » » 4 years ago, # ^ |   0
 » 4 years ago, # |   +2 why it cannot be greater than (a+b)/3 in problem A.
•  » » 4 years ago, # ^ | ← Rev. 2 →   -19 this comment is deleted because of negative feed back
•  » » 4 years ago, # ^ |   +14 Just for simple realisation :We need three objects in total to make anything(2 of first and 1 of second kind and vice versa). Now if each object need 3 material then you can create at max (a+b)/3 objects
•  » » » 4 years ago, # ^ |   +6 got it now :)
•  » » 4 years ago, # ^ | ← Rev. 4 →   +8 number of 1-2 is x, 2-1 is y then x + 2y <= a 2x + y <= b -> x <= a, y <= b, x + y <= (a + b) / 3
•  » » 4 years ago, # ^ | ← Rev. 3 →   +9 Notice that we the answer actually asks how many times we can subtract 3 from (a+b). Obviously the answer is (a+b)/3. But for some restriction on the type of objects we had to go for min(a,b,(a+b)/3). There is another way. Suppose we take X diamonds and 2X sticks to build X shovel. And we take Y sticks and 2Y diamonds to build Y sword.Total number of used items = 2X + X + 2Y +Y = 3X +3Y. So,(a+b) - (3X+3Y) = 0, why right hand side = 0 ? Because if we want to maximize the answer, number of items left will be zero. Solving for X+Y gives (a+b)/3.
•  » » 4 years ago, # ^ |   +3 lets assume we will be able to make x shovels and y swords. 1. 2*x+y<=a 2. x+2*y<=badding above equations we will get 3x+3y<=(a+b)hence (x+y)<=(a+b)/3
 » 4 years ago, # |   0 Ah, finally the editorial is out now.
 » 4 years ago, # | ← Rev. 2 →   0 We can do D in this way too -1 ) Make prime sieve (sieve of Eratosthenes) to check a no. is prime or not2 ) Convert given array to set3 ) Now if a value in set is prime then store (-1,-1) else do brute force to check each valid pair of the divisor.It passes better than SPF algo (in my case)Here is link : My_Solution
•  » » 4 years ago, # ^ |   +3 btw the algo is good ..btt your submission is completely messyy !
•  » » 4 years ago, # ^ |   0 Nice approach to avoid duplicate numbers. For most numbers brute force works fast enough. Or may be test cases are weak. But I am not sure because some sophisticate time complexity analysis is needed.
 » 4 years ago, # |   0 can i get solution for problem D in c++??
•  » » 4 years ago, # ^ |   +3 You can refer this 83534816
•  » » » 4 years ago, # ^ |   +3 hey i like your solution. its nice
•  » » 4 years ago, # ^ |   +1 I you click the contest page you see these numbers right of the problems. Clicking them will show you a list of that problems submissions.
 » 4 years ago, # |   +6 This could be A-hard version :) 478-C Table decorations
 » 4 years ago, # |   +3 Hey awoo, I enjoyed the contest, interesting questions! I was wondering why you chose to put F and G in this contest: aren't they better suited for a div 1?
•  » » 4 years ago, # ^ |   +4 Bro, this round is called an "Educational Round" right? So it seems reasonable that there will be some problems with advanced techniques at the high end of the problemset.
•  » » » 4 years ago, # ^ |   -6 getting tle on case 46 for E on this......Can anyone tell why?
•  » » » » 4 years ago, # ^ | ← Rev. 3 →   +11 AbhishekAg your solution uses an unordered map, change it to an ordered map (normal map) and it will pass. Check this : https://codeforces.com/blog/entry/62393One more thing I must say is that I have observed you have commented the same thing you commented here atleast 3 times as an sudden, out of nowhere reply to various comment threads here, such spamming is really inappropriate and irritating! Create a new comment if you need to.Please don't do this.
•  » » » » » 4 years ago, # ^ |   0 Sorry for spamming, I thought I could delete all comments once I get a reply......but it seems that cannot be done.....I keep it in mind from next time... Also.....can you please explain why changing to map got accepted?
•  » » » » » » 4 years ago, # ^ |   +8 Check the link I shared, its basic idea is that on average unordered map has O(1) operations, which is OK if inputs are random, but it turns out, if someone wants, they can generate an adversarial test case to cause an O(n^2) blowup on it.
•  » » » » » » » 4 years ago, # ^ |   0 thanks...
 » 4 years ago, # | ← Rev. 2 →   +5 Alternative for D, My solution for D is little bit different https://codeforces.com/contest/1366/submission/83482692 It uses the observation that if (d1 and d2) solution exist for a number A such that gcd(d1+d2,A) =1.Then it will definitely exist in the form of d1 = x, d2 = A/x where x is some divisor of number A.
 » 4 years ago, # |   +38 Construction for A: let $K = \min(A, B, \frac{A + B}{3})$. Each of the $K$ items crafted needs at least one stick and at least one diamond. Set those sticks and diamonds aside; it's possible since $K \leq A$ and $K \leq B$. Each item then needs one additional resource of either type. We have $A + B - 2K$ resources remaining. It follows from $K \leq \frac{A+B}{3}$ that $A + B - 2K \geq K$.
•  » » 4 years ago, # ^ | ← Rev. 2 →   -24 getting tle on case 46 for E on this......Can anyone tell why? upd : solved
 » 4 years ago, # |   0 Can someone explain the time complexity of Problem D?
•  » » 4 years ago, # ^ |   +3 Let A be max Number which is 107 and n is size of array -To precompute SPF (smallest prime factor ) Time complexity is O(AloglogA)Now for each value we can answer in O(logA)As there are n values => O(nlogA)Total = AloglogA + nlogA
•  » » » 4 years ago, # ^ |   0 How is the complexity of SPF nloglogn ?
•  » » » » 4 years ago, # ^ |   0 Complexity of sieve of eratosthenes is O(nloglogn)
•  » » 4 years ago, # ^ |   +7 You can solve Problem D in O(n+A). int n; int p[10000005], ptop; int pk[10000005]; void sieve() { int n = 10000000; for(int i = 2; i <= n; ++i) { if(!pk[i]) { p[++ptop] = i; pk[i] = i; } for(int j = 1; j <= ptop; ++j) { int t = i * p[j]; if(t > n) break; if(i % p[j]) { pk[t] = pk[p[j]]; } else { pk[t] = pk[i] * p[j]; break; } } } } int ans1[500005]; int ans2[500005]; void solve() { sieve(); scanf("%d", &n); for(int i = 1, x; i <= n; ++i) { scanf("%d", &x); if(x == pk[x]) { ans1[i] = -1; ans2[i] = -1; } else { ans1[i] = pk[x]; ans2[i] = x / pk[x]; } } for(int i = 1; i <= n; ++i) printf("%d%c", ans1[i], " \n"[i == n]); for(int i = 1; i <= n; ++i) printf("%d%c", ans2[i], " \n"[i == n]); return; } This code spend 389ms to solve Problem D without any fast I/O.To precompute some number's smallest prime factor, using The sieve of Euler, time complexity is O(A).But it is not necessary to divided by smallest prime factor one by one, which using O(log(A)).You can precompute number x has how many smallest prime factor at the same time.
•  » » » 4 years ago, # ^ |   +31 put your code in spoiler or provide link.
•  » » » 4 years ago, # ^ |   0 Can u please explain how the complexity of your sieve ( which is euler sieve ,I don't know about this please if some tutorial then give ) is O(A) ?
•  » » » » 4 years ago, # ^ |   0
 » 4 years ago, # |   +3 Can anybody explain a little on why are we traversing the arrays in reverse order helps?
 » 4 years ago, # |   0
 » 4 years ago, # | ← Rev. 2 →   +4 Proof for A:It is evident, as the tutorial mentions, that it cannot be greater than min(A,B,(A+B)/3), as you said. This does not imply that we can always achieve one of these optimal cases! For the first two, it's easy to see that if A>2*B, then it is optimal to pair each B with 2 A's and we have A's to spare. WLOG for B>2*A. For the other case, we can take 2 of the one with greater quantity and pair it with 1 of lesser quantity. It is easily observed that this will reduce max(A,B) by 1 more than it reduces min(A,B), so we will always be bringing the values A and B closer together, except when they are equal, in that case we create a difference of one. By the above two observations, this process ends either when one of the piles are 0, in which case the other must be 1, or when both piles are 1. This corresponds to cases (A+B)%3=1 and (A+B)%3=2 respectively. Hence we complete the proof that not only is the optimal less than the 3 constraints, but there exists a way to achieve these 3 constraints.
 » 4 years ago, # |   0 liked the problems a lot, thank u!
 » 4 years ago, # |   0 Please help! In C, why do we check if i <= j?
•  » » 4 years ago, # ^ |   0 Let's say the count array is of length 7, odd length Now what we equate b/w is: (cnt[0],cnt[6]) (cnt[1],cnt[5]) (cnt[2],cnt[4]) (cnt[3],cnt[3]) Notice that, equating between cnt[3] and cnt[3] case (i == j) i.e, when both pointers meet at mid is unnecessary and so is, any index > 3 case (i > j) as we will be repeating the process(think).Case (i == j) won't happen for even length count array.
 » 4 years ago, # |   0 Can anyone help me find the complexity of my solution https://codeforces.com/contest/1366/submission/83570373
•  » » 4 years ago, # ^ | ← Rev. 2 →   0 it's Nlog(log(N)) where N = mx (in your code).
 » 4 years ago, # |   0 Can anyone explain the proof of how to choose d1 and d2 effectively in Problem D? The editorial version is not that clear to me.
•  » » 4 years ago, # ^ |   +6 Simple ApproachFor an $even$ number, answer will be $($ $2$, Product of remaining odd prime factors $)$For an $odd$ number, answer will be $($ $1st$ Smallest prime factor, $2nd$ Smallest Prime factor $)$And obviously, first, you need to check whether alteast $2$ distinct prime factors for a number exists or not. if not answer will be $($ $-1$, $-1$ $)$ProofFor an $odd$ number, Consider an example $ai$ = $105$ $( 3 * 5 * 7 )$. Ans is $(3, 5)$.$3$ is $1st$ smallest prime factor and $5$ is $2nd$ smallest prime factor of $105$.Let $x = d1 + d2 = 3 + 5 = 8$.$g = gcd(x, 105)$ and obviously $g$ can't be $3$ or $5$. So $g$ should be greater than $5$, which is not possible. (why? Let $x' = g * e$ , $e$ is even number, $e$ must be aleast $2$. You can see $x' > x$ if $g > 5$, which is not possible.So $g$ has to be $1$.For an $even$ number, Consider an example $ai$ = $210$ $( 2 * 3 * 5 * 7 )$. Ans is $(2, 105)$.$105 = 3 * 5 * 7$ (Product of remaining odd prime factors).You can see $d1 = 2$ and $d2 = 105$, now forget about $d1$ and ask a question from yourself. What is the minimum $y$, I should add to $d2$ such that $g = gcd(d2 + y, ai) > 1$. And you will find you need to add smallest prime odd factor, for this case it is $3$ but we are adding just $2$ ($d1 = 2$, hence the answer).
 » 4 years ago, # |   0 For the code for problem G proiveded in the editorial, can anyone please explain why the following transition is skipped:-dp[i+1][j-1]=max(dp[i+1][j-1],dp[i][j]), where the ith character of the first string is a '.'.I am not able to convince myself why this transition is redundant.
•  » » 2 years ago, # ^ |   0 Yeah, so the transition look like this: TAKE character because it will be in *final string DONT TAKE character because it won't be in *final string, delete it DONT TAKE character because it won't be in *final string, but we must destroy it with some character '.' *final string I am talk about is string t we should get from s after applying function f.The transition you are talking about cannot ever happen, because if we took some character with transition 1. , by definition it will be in final string so it shouldn't be destroyed.
 » 4 years ago, # |   +6
•  » » 4 years ago, # ^ |   +1 Try map instead of unordered_map.
•  » » » 4 years ago, # ^ |   0 But unordered_map is faster than map?
•  » » » » 4 years ago, # ^ |   +11 No , unordered_map will take linear time in worst case.for more details refer this
•  » » » » » 4 years ago, # ^ |   0 Ohk, I got that , can you tell me when unordered_map gives us answer in O(1) with surety , and when it can take linear time?
•  » » » » » » 4 years ago, # ^ |   +1 Sorry but I don't know that much deep.
•  » » » » » » » 4 years ago, # ^ |   0 How, would i know where to use map or where to use unordered_map?
•  » » » » » » » » 4 years ago, # ^ |   0 Times you need to pay to insert or read in map is O(logn),while on unorederd_map maybe O(1) in average but in O(n) in some test data. So if you are sure that time for you is enough to use map, just use it.
•  » » » » » » 4 years ago, # ^ |   +1 The principle of unordered_map is hash. Maybe you need to learn about this, for the random test data it cost O(1) to insert, but sometimes it will cost O(n) to insert.
•  » » » » » » » 4 years ago, # ^ |   0 Yes,I know about hashing and collisions but it was also taught that inbuilt hash function is very good and chances of getting O(n) complexity in unordered_map is very less , how would I identify by looking the constraints in the question that unordered_map will give linear time in this case?Plz help
•  » » » » » » » » 4 years ago, # ^ | ← Rev. 3 →   0 You can write hash by yourself and get the module randlomly because the moudle in unordered_map is a const number (I guess), and in C++ you may write "int mod=rand()" or in any other ways. In this way you may make sure that the cost is O(1) because the test data don't know what you moudle is because it's created randomly. But if it is possible I think you'd better use map because it is more convenient.
 » 4 years ago, # |   0 Why i am getting TLE for D.i think it's complexity is same as described in editorial?https://codeforces.com/contest/1366/submission/83593262
•  » » 4 years ago, # ^ | ← Rev. 2 →   +3 you are applying factorization every time you are taking an input,the whole point of storing the minimum prime using seive is to not do the thing which you are doing and get it done in less time so your code takes O(sqrt(n)/log(sqrt(n))*log(n))) every time and we once applied seive we can get it's prime factorization in O(log(n)) time
•  » » » 4 years ago, # ^ |   0 Not Sure about Complexity you described.I thought it's log(ai).Can u provide some source that prove your Comlexity(O(sqrt(n)/log(sqrt(n))*log(n))))!!TIA
•  » » » » 4 years ago, # ^ |   +3
 » 4 years ago, # |   0 I have a conceptual doubt in problem D.Let's say our number N has its prime factors p1,p2,p3,p4.if we choose d1=p1 and d2=p2.p3.p4 ,then IS it NOT possible that x=p1.p3(let) can produce (d1+d2)modx=0 ?since d1modx!=0 and also d2modx!=0.
•  » » 4 years ago, # ^ |   +3 d1 = p1 and d2 = p2 p3 p4d2 makes it so d1 can never be p1 mod x = 0but also d1 makes it so d2 can never be p2 mod x = 0. think about it this way, think about d2 as a multiple of p2. a multiple of p2 + p1 can never be mod x = 0. repeat for all prime factors and yeah
•  » » » 4 years ago, # ^ |   0 your comment encouraged me to pick up pen and paper and disprove myself wrong by writing just 2-3 lines using basic arithmetic modulo.so thanks a lot :)
 » 4 years ago, # | ← Rev. 3 →   0 Problem E is very interesting. Can someone please help me figure out the best possible solution for problem E if condition b[i] < b[i+1] was not in place.
 » 4 years ago, # |   0 Getting wrong answer in the 2nd testcase of C. Can someone please help me find the mistake? Here's the link to my solution: 83472337
•  » » 4 years ago, # ^ |   0 15 21 00 11 10 11 1try this testcase ...
 » 4 years ago, # |   0 can anyone help me ??what's wrong in my implementation of problem D?
 » 4 years ago, # |   0 What was the intended purpose of lst < prv in the model solution for F? I removed this check and the code is still accepted.
 » 5 months ago, # |   0 Problem A is just bad.
•  » » 4 months ago, # ^ |   0 Maybe, just maybe... you're bad?
•  » » » 4 months ago, # ^ |   0 That too yeah...