### chokudai's blog

By chokudai, history, 2 years ago, We will hold AtCoder Beginner Contest 138.

The point values will be 100-200-300-400-500-600.

We are looking forward to your participation! Comments (81)
 » Oh~All of us can participate in the Codeforces contest after finishing the Atcoder Beginner Contest,that's so good!Good luck to all && High rating :D
•  » » Agreed. Wish High Rating!
 » (Unofficial) I'll do English problem discussion streaming starting from 13:50 UTC (10 minutes after the contest): URLIf you are interested but you are participating in the CF round, check it after the contest!
•  » » Will you do it for codeforces (div1 ?) too?
•  » » » No, it's too late here
•  » » » Tu div 1 kabse hogya patel ?
 » Problem statements on AtCoder be like •  » » when you finished the contest in less than 30 min. so you make a meme in free time during the contest
•  » » » And Cheetas like you there to reply
•  » » » » i am getting TLE on E problem any solution i am doing it in 1 traversal dont know why ?? please help
•  » » » » » my first 32 test cases out of 40 get passed & rest of them are giving WA:'(
•  » » » » » » Maybe you did not use long long?
•  » » » » » » » no long-long in Python :(
•  » » » » when you first time give AtCoder & get rank under 500....#Durgesh the coder
•  » » » » » same here i have used binary search as well then 40 passed but 8 are giving TLE i dont know why ?? anyone know the solution please share
•  » » » » » » yup use c++ i was using python in c++ it passed @chaudhary use c++ it will be passed and do the binary search . not using two loops hope it helps ?
•  » » » » » » » I was also using python & binary search....sed lyf :'(
•  » » » Actually, I didn't participate in the contest.
•  » » In Atcoder infinity is always an even number
 » ABCDE is quite easy, F is so hard.
•  » » getting TLE in both D and E ;(
•  » » » 2 years ago, # ^ | ← Rev. 4 →   To avoid tle in D you can use path prefix sums and do dfs just once.
•  » » » » This is exactly what I did...
•  » » You are Purple ...but question E is hard for a noob like me
•  » » » 2 years ago, # ^ | ← Rev. 2 →   E is kinda classical problem, thus is easy to those who have solved similar problemPS: F is hard XD
•  » » » » Could you give me a link to a problem that is similar to E on a education oriented site. Thanks
 » 2 years ago, # | ← Rev. 2 →   I didn't realized that the contest is not ended yet.
•  » » even if it is you shouldn't publicize it in midst of contest
•  » » Don't you think it would have made more sense to post this comment after the contest ends? I think you should delete the comment so that others won't gain an unfair advantage.
 » How to solve F? I was thinking with digit dp but was unable to think the states. Please Help
 » How to solve problem F?
 » Quick editorial: A: Straightforward implementation. B: Straightforward implementation. C: The answer is a linear combination of the ingredient values. The earlier an ingredient is inserted into the pot, the smaller its coefficient will be in the final sum. So, they should be inserted in non-decreasing order of their value. D: Accumulate $s_j$ = sum of $x$ over all operations on vertex $j$. The task is to compute for each vertex the sum of $s_j$ over all of its ancestors (including itself). We can do it with a DFS on the tree. E: Process $s$ into sorted lists $\ell(c) =$ the indices at which character $c$ appears in $s$. We can use binary searches on these lists to compute $f(c, i) =$ the first index in $s'$ greater than $i$ at which character $c$ occurs. The answer is $f(t[-1], f(t[-2], f(t[-3], \dots f(t, f(t, -1)))))$. F: Notice that for $y \geq x$, $(y \textrm{ XOR } x) \geq y - x$ and $(y \textrm{ MOD } x) \leq y - x$. Hence we need $(y \textrm{ XOR } x) = (y \textrm{ MOD } x) = y - x$. These constraints mean that if we construct the bits of $x$ and $y$ from most significant to least significant, they should share their leading bit and the set of "on" bits in $x$ should be a subset of the "on" bits in $y$. We can do a DP to construct pairs of digit strings $x, y \in [L, R]$ having those properties with state: (number of digits placed, has $x$ diverged from $L$, has $y$ diverged from $R$, has leading digit been placed). The implementation is quite short and perhaps easier to understand.
•  » » maybe easier to understand without your template...
•  » » Thanks for the quick explanations. In F, could you please explain a little more on These constraints mean that if we construct the bits of x and y from most significant to least significant, they should share their leading bit and the set of "on" bits in x should be a subset of the "on" bits in y. I mean how do we infer this? Also what exactly does this mean the set of "on" bits in x should be a subset of the "on" bits in y.. An example would really help! Thanks in advance!
•  » » » 2 years ago, # ^ | ← Rev. 3 →   We need $y \geq x$ with $(y \textrm{ XOR } x) = (y \textrm{ MOD } x) = y - x$. $(y \textrm{ XOR } x) = y - x$ is true when $y$'s binary representation has $1$'s in all of the places that $x$'s binary representation does. $(y \textrm{ MOD } x) = y - x$ is true when $x > y/2$, which (in combination with the other requirement) simply means that they should have the same leading bit.
•  » » Hi can you tell what is wrong in my approach I am using DP for problem C just like in matrix chain multiplication.Here is my codecodeI was able to solve D and E but not C can you please help me..
•  » » » 2 years ago, # ^ | ← Rev. 2 →   I don't think there's a need for dp here.I used a greedy approach which is to build a multiset or priority queue and we just can take the minimum $2$ elements and remove them from the queue and push the new value $(x+y)/2$ to the queue until we end up with one element only which wil be the answer.
•  » » » the main logic behind taking the smallest element first is that every time you take two number and add it's sum to by taking half to another number the relevance of first number to your new sum becomes 1/4 then 1/8 then 1/16 .......and so on. that's why u would always like to decrease the smallest u have and add the largest numnber to your answer as here u want the sum to be maximum if they would have asked for minimum u would have gone another way round.that's why the approach of using priority_queue, multiset or sorting works. I hope this helps
•  » » » » I understand that but why does dp give wrong answer??
•  » » » I don't think your solution is not standard, this problem just sorts arrays and solve
•  » » » » You are using false negatives, you could have simply said that my solution is not standard but now it seems as if you mean to say that my solution is standard ??
 » even though i know memset can fuck me up at times, why do i still use it :| ps:F anyone ?
 » How to solve F,it seems to be a math problem
 » How to solve F?
 » How to solve E and F?
»
2 years ago, # |
Rev. 2

For Task E, I used Binary Search just like others. But still I am getting TLE in 4 test cases even in C++. I tried to think really hard, but could not find the problem. I would really appreciate if someone can help me in this regard.

~~~~~

# include <bits/stdc++.h>

using namespace std;

# define ll long long

int main() {

string s,t; cin>>s>>t;

ll n = s.length(), m = t.length();

vector<vector > ids(26,vector());

for(ll i=0;i<n;i++) { ids[s[i]-'a'].push_back(i); }

ll prev = -1, ans=-1, count=0;

for(ll i=0;i<m;i++) {

char a = t[i];

if(!ids[a-'a'].size()) {
cout<<-1<<endl;
return 0;
}
else {
vector<ll> curr = ids[a-'a'];

ll low = 0, high = ((ll)curr.size())-1, res=-1;
// cout<<1<<endl;
if(prev<=curr)
res = curr;
else if(prev>curr[high])
res = -1;
else {
while(low<=high) {
ll mid = low+(high-low)/2;
if(curr[mid]==prev){
res = curr[mid];
break;
}
else if(curr[mid]<prev) {
low = mid+1;
res = curr[low];
}
else {
high = mid-1;
}
}
}

if(res==-1) {
count++;
ans = count*n + curr;
prev = curr+1;
}
else{
ans = count*n + res;
prev = res+1;
}
}

} cout<<ans+1<<endl; return 0; } ~~~~

•  » » 2 years ago, # ^ | ← Rev. 2 →   I think its because of vector cur = ids[a -'a'] if there is 10 ^ 5 elements in ids[a — 'a'] and copying it 10 ^ 5 times from ids[a — 'a'] will be TLE, i think you can binary search for just ids[a — 'a'] without creating new array
•  » » » Thank You so much. If you would not have told this, I would have been still wondering and be in doubt. Thank You really.
 » For problem E I have thought of an idea using like most have said binary search but i AM GETTING RE and I don't know whyCan somebody please help me understand my mistake
 » 2 years ago, # | ← Rev. 2 →   For me same problem with E.I still don't get it why it does not work. I inspected several solutions. They all do more or less the same. But mine does not work :/Can somebody point me in the right direction, or, once again post the link to the testcases? Thanks!https://atcoder.jp/contests/abc138/submissions/7001559PS: And I think that code looks really nice, fairly simple solution ;)
•  » » 2 years ago, # ^ | ← Rev. 2 →   You are doing upper_bound every timeLet's check this test case s = contesc, t = cin line 53 you take ll ans2 = 0, and in this test case your answer would be 7 instead of 1You need to take ans2 as -1 in the start
•  » » » OMG... Thank you.
 » Will test cases be made available?
•  » » Test cases are usually published weeks to two months later.
 » Problem E can be solved without binary search. My sub
•  » » 2 years ago, # ^ | ← Rev. 2 →   This is O(n^2). Looks like testcases are weak chokudai Testcaseaaaaaaaaaaaaaaaaaaa..........aaaab (a 99999 times then b)bababababa......................ba (alternating ba)
 » where editorials can be found for this contest??
•  » » You have to switch to japanese version in order to find it
•  » » » ok thanks
•  » » » how can i get Japanese editorials?
•  » » » »
 » Can any one give tutorial for Problem D and E
 » Some test cases were added after the contest for problem D to deal with particular cases which vertex ai is not a parent of bi. So some codes on AC list are not useful anymore.e.g. V = {1, 2, 3}, E = {1-2, 2-3}
 » 2 years ago, # | ← Rev. 2 →   What's wrong is this submission for Problem E ?? https://atcoder.jp/contests/abc138/submissions/7016567 Thank you.
 » can anyone please explain logic behind Question D . It will be very helpful
•  » » simply update Q queries and update counter of all adjacent vertices including x using dfs or bfs.
 » In question D, my BFS solution got AC but now it is failing for 3 test cases added after competition, whereas DFS is getting AC unable to find the glitch, please help. My DFS and BFS codes are given below: //DFS code: #include #include #include // #include // #include #include using namespace std; typedef vector vi; typedef vector vvi; void dfs(int v, int p, long long cost, vvi &a, vector &c, vector &up){ c[v] = cost+up[v]; for(auto u: a[v]){ if(u==p) continue; dfs(u, v, c[v], a, c, up); } return; } int main(){ int n,m; cin>>n>>m; vvi a(n+1); for(int i=0;i>x>>y; a[x].push_back(y); a[y].push_back(x); } vector up(n+1, 0); for(int i=0;i>x>>y; up[x] +=y; } vector c(n+1, 0); dfs(1, -1, 0, a, c, up); for(int i=1;i #include #include #include #include #include using namespace std; int main(){ int n,Q; cin>>n>>Q; vector> adj(n+1); for(int i=1;i>p>>c; adj[p].push_back(c); } vector cost(n+1,0); for(int i=0;i>v>>c; cost[v] += c; } queue q; int a = 0; for(int i=0;i0){ a=i; break; } q.push(a); // vector vis(n+1,0); // vis[a] = 1; while(!q.empty()){ int nc = q.size(); while(nc>0){ int pa = q.front(); q.pop(); nc--; long long pc = cost[pa]; for(int i=0;i
•  » » Does your BFS solution fail because of Time Limit? Or Wrong Answer?
•  » » » 2 years ago, # ^ | ← Rev. 3 →   Wrong answer. I understood the problem with my bfs code. The tree is undirected and rooted at 1. So, in my adjacency list I should push both x-y and y-x and keep track of visited. Updated solution with BFS. #include #include #include #include #include #include using namespace std; int main(){ int n,Q; cin>>n>>Q; vector> adj(n+1); for(int i=1;i>p>>c; adj[p].push_back(c); adj[c].push_back(p); } vector cost(n+1,0); for(int i=0;i>v>>c; cost[v] += c; } queue q; q.push(1); vector vis(n+1,0); vis = 1; while(!q.empty()){ int nc = q.size(); while(nc>0){ int pa = q.front(); q.pop(); nc--; long long pc = cost[pa]; for(int i=0;i
•  » » 2 years ago, # ^ | ← Rev. 3 →   In DFS code, you linked both (a[x], a[y]) and (a[y], a[x]), though only (a[x], a[y]) in BFS one.Additional testcases represent a tree structure like "1-3-2" (V = {1, 2, 3}, E = {(1, 2), (2, 3)}), which x isn't a parent of y.Try the case below. AC is "1 3 1" and DFS returns it but your BFS return "1 2 1".3 21 32 31 12 2
•  » » » Thanks I just realized that mistake.
•  » » » Thank you, I finally know where I am wrong, I thought the data was given by the structure of the tree....
 » For D, I actually used dfs timer to convert the tree into an array, then did the prefix sum trick.
•  » » 2 years ago, # ^ | ← Rev. 2 →   Can you provide some resources? for the problem...
 » Can anybody please tell me where are the editorials?
•  » » You can find in Japanese mode or access directly to the link below. https://img.atcoder.jp/abc138/editorial.pdf
 » 2 years ago, # | ← Rev. 2 →   They should provide editorials in English, how to solve d?
•  » » For this time, there's English edition from p.7 https://img.atcoder.jp/abc138/editorial.pdf
 » Is there tutorial part for Atcoder Problem ?
 » Hi, I'm getting WA with This submission, any help?
 » I need help in problem D.I got tle error My submission