We will hold Toyota Programming Contest 2024#7（AtCoder Beginner Contest 362）.

- Contest URL: https://atcoder.jp/contests/abc362
- Start Time: http://www.timeanddate.com/worldclock/fixedtime.html?iso=20240713T2100&p1=248
- Duration: 100 minutes
- Writer: sotanishy, toam, ynymxiaolongbao, physics0523
- Tester: MMNMM, nok0
- Rated range: ~ 1999
- The point values: 100-200-350-425-475-550-575

We are looking forward to your participation!

Can you give the topics for each question

A-Brute Force

B-G- Segment Tree and Binary search problems.

Good luck, my friends!

you,too!

GL&HF!

..

Liked F but G ruined everything

G btw

F btw

Damn that's just sad now

But it doesn't demand a certificate.

This is very similar too.

G btw

how did u do F was it Hungarian?

My logic:

Let the root of the tree be 1, you will notice that the most optinal way to the make the matchings is to make pairs between nodes such that they are in subtrees of different children of 1.How you make the matching doesn't matter since the total weight of the matching=sum of distances of all nodes from root i.e. 1.

Now, if any children of the root of the tree i.e. 1 have subtree size > $$$n/2$$$ where n is number of vertices in tree. You can't make matchings for all the nodes in the optimal way mentioned above. To overcome this, reroot the tree such that the subtree size of all the children of the root is <= $$$n/2$$$. This root always exists and can be found using dfs.

Now, To implement the construction of the matching there are many ways. What i did was make a set of the subtree sizes of children of the root and then take a node from both the subtree with smallest and largest size and pair them up, it doesn't matter what node you take as long as each node is taken once. Do this till all nodes have been matched.

I just find that I have almost the same idea as yours (but failed finishing it during contest), thank you so much for sharing your solution! By the way, in the step of "reroot the tree ...", is the new root just the centroid of the tree? (I think it is but not quite sure)

Yes, it is the centroid

E took my energy

Are you kidding? Why is task G the template of the Aho-Corasick Automaton?

Here is the template task in Luogu. They are the same.

and like this: https://vjudge.net/problem/LightOJ-1427

Is g some standard problem?

Yes, very standard. Either Aho or Suffix Array will do it.

yep its a cses problem

Yes

How could you not know that with that username

I know trie but not aho corasik algo

Please stop giving problems that can be solved by just copying the template like G.

how to solve E

very ez that dp

I did it using 3d dp.

This is my submission :https://atcoder.jp/contests/abc362/submissions/55562909 (was my first contest at atcoder :))

I set dp[i][length][diff] = number of such sequences such that the ending position of subsequence is i , the length of subsequence is length and the difference of the AP is diff , i kept diff in a dictionary(map) as the value of diff could reach upto 10^9.

After that just used dp[i][length+1][diff] += dp[j][length][diff]

Since if there is subsequence of length k with difference diff and the current element minus j th element is equal to diff , then we could make a new subsequence of length legth+1

Then for the final answer , just printed the summation of length for all i and diff.

How do you solve C? I managed to solve D and E but couldn't solve C

The minimun sum u can construct is sum of all l[i] and maximun is sum of all r[i]

lmfao I wish I skipped C then coz by the time I finished it I had no time for either D or E

Yes,me too.D is a standard problem and E is a DP problem.But C is harder than D.

To construct the Xs you can initialize them to the maximum sum they can give, then decrease them until the sum is 0. You can also probably go in the reverse direction as well, start with minimum sum and increase

Can't believe so many people would SAM.

In fact,ACAM is enough.

Perhaps a more appropriate statement would be "KMP on Trie".

Also can't believe so many people would ACAM. ==

You can also use ACAM to solve G.

Most of them just copy the others' code. There is a similar problem on a Chinese OJ and there are published solutions with code.

Atcoder, if you aren't able to set Gs then just remove them.

Solved F a couple of minutes after the contest has ended, eternal pain and suffering

Should have taken a hint from the problem name. Btw what's the actual way to do a matching in a case like this, without bruteforcing and abusing move-semantics?

Problems such as problem G shouldn't appeared in the contest. It is too original, and you just need to copy and paste without any thinking. Even worse to make G 575 but F 550, it should be reversed. It's super unfair for candidates who can solve F but doesn't know anything about AC automation.

Don't know ACAM means too weak. Just practise more.

how to do F ?

I could get somewhat idea to use Hungarian but that O(n^3) then whats the way to solve F or am i overthinking and using something which is useless like Hungarian in this case

I kind of guessed F, if you find a centroid and start matching nodes from different subtrees of centroid then the actual pairs do not matter since they won't affect the answer, ie sum of all distances should stay the same (at least it feels like it). Then you have reduced the problem to matching integer values from different sets.

I don't like this round. It has a template problem like G, it makes this round boring and unfair.(just my opinion)

Solved E after 3 mins the contest ended... Fortune enough to notice G is suffix array to cover the loss in E.

Can you tell me why this code fails

I have a counter case I am not able to figure out where I am wrong 5 1 1 1 1 1

(Code)

E is similar to this problem.

But their ideas are different.

The G problem is not a original problem.

Delete G.

Can anyone help I am getting tle on last 2 test cases of problem d int main(){ ios_base::sync_with_stdio(false); cin.tie(NULL); cout.tie(NULL); lli n,m; cin>>n>>m; vectorarr; arr.push_back(0); for(int i=0;i<n;i++){ lli x; cin>>x; arr.push_back(x); } vector<pair<lli,lli>>mp[n+1];

for(int i=0;i<m;i++){ lli x,y,z; cin>>x>>y>>z; mp[x].push_back({y,z}); mp[y].push_back({x,z}); } vectordis(n+1,1e18); dis[1]=0;

priority_queue<pair<lli,lli>, vector<pair<lli,lli>>, greater<>> pq; pq.push({arr[1],1}); while(pq.size()>0){ lli dd=pq.top().first; lli vv=pq.top().second; pq.pop(); for(auto it:mp[vv]){

} dis[1]=0; for(int i=2;i<=n;i++){ cout<<dis[i]<<" "; }

return 0; }

bhai, you are pushing same node multiple times in the PQ.

Lets say, graph,

lets say, the input is

Please check, how many times you are pushing node value '2' in the queue.

Yeah I got my mistake,thanks a lot

You should create an array called

`bool vis[]`

which stores whether each node has already been processed. Without this array, sometimes a node can be processed multiple times. Although this doesn't result in an infinite loop, this still significantly adds to the execution time. (Processed multiple times means the neighbors of the node is relaxed multiple times, which isn't necessary.)FYI the solution to this problem is that for every edge

`u->v`

with weight`w`

, we simply add`arr[v]`

to`w`

. Then, we run a standard dijkstra and add`arr[1]`

to our answer. Proof of correctness is fairly straightforward.PS: It is recommended to send submission links rather than the whole code. This is not only more aesthetically pleasing but also gives us more information on the submission, whether it TLEs or WAs, etc.

Thanks I got my mistake, yes I'll paste the link next time onwards

can you share the submission link using visited array with this code ?

u can check the submission here the mistake I was doing is that some there will be a lot of overhead so to prevent that you can add a condition if distance at that place is less than dd than just continue and avoid visiting it's neighbours at it would only increase the cost ...

yeah, it works same as using the visited array before checking for it's neighbours, if the node is visited then continue

I copied pasted my solution to G from CSES and changed 0 lines. I don't think problems like this should appear in a rated round. If problems like this are to appear, it should be in an unrated educational round (similar to the Atcoder DP contest a long time ago).

Link: https://cses.fi/problemset/task/2103

[DELETED]

This round was particularly frustrating, because I haven't seen G before and therefore spent a substantial amount of time on G. Hence, I finished the contest with ABCDE_G. However, I realized the greedy solution for F merely 5 minutes after the contest. If I had seen G before, I could focus on F and would have definitely full-solved. Not only did I not full-solve, my rating is probably going to be affected too, due to the fact that so many people have seen G before. I think many people are having the same thoughts, so atcoder_official, if you can't set G's, then don't. Having a contest with 6 problems is better than having a template question as the highest-point problem.

Can you please check this submission ? why is this getting WA, just can't figure out. Can't figure out the error :| .

https://atcoder.jp/contests/abc362/submissions/55580896

jatrophalouvre

Breakcase:

Your code gives:

Correct answer:

thanks a lot.

In fact, problem F is so similar to 1387B2 - Village (Maximum).

can someone please help me figure out the mistake in this ?

https://atcoder.jp/contests/abc362/submissions/55580896

Does Dijkstra with Min Heap priority queue not work for anyone for D? It gives TLE for me. So is the solution using Fib Heap? Or is it some optimization issue?

It worked for me with just some slight modification - At line 52, 53 and 69, 71 --> rest the whole algorithm will be normal djikstra

You can check — My submission

Ohh it looks very similar to my approach too!

Can you check this submission out: https://atcoder.jp/contests/abc362/submissions/55586366

I'm not sure of the TLE in this. The approach is more or less the same. Also in my previous version of reply I had a silly mistake of not including V,U edge which caused WA

Inside the solve function --> you are creating a self loop => which may cause both TLE and Wrong answer

graph[u].push_back(Edge(v, b)); graph[u].push_back(Edge(u, b)); --> v must be there

I think it should get accepted, feeling bad for the typo..

Lol yea saw that. That was silly haha. Still TLE tho.

Edit: Turns out my visited array I was using for not doing repetitive computation was not doing the trick. I should've used the dist array itself. Got it now. Thanks for helping me out bro!

Problem G is a template of Aho–Corasick algorithm.

By the way, it's not Aho–Corasick algorithm but Aho–Corasick Automaton.

But it's such a name in Wikipedia.

Well, you see,Aho-Corasick is not a Algorithm but a data structure based on trie, so I think there might be an error in Wikipedia.

Yes you are right, its name is indeed this.

Hello there!. Please help me with this submission why TLE!!

https://atcoder.jp/contests/abc362/submissions/55595549

Hi!You should add

`if (vis[u]) continue;`

to 145th line.Yeah I got it. Thanks

But why exactly this works. You can see that I am checking for vis while traversing for all the neighbors.

Possible reasons for this are that a point's neighbor is placed on the back end of the priority queue, causing the point to not be marked in time, and it can also cause a node to queue many times and time out.

Ok I see. Traversing the neighbors of the same node more then once is a problem. If the node enters in the pq then it will get it neighbors checked this may happen multiple times. Now I get it.

Thanks mate!!. (Mentioning this to not follow the same in future).

Can anyone please spot the mistake in my code for C question I am getting 2 Wrong answers rest all accepted.

My submission: https://atcoder.jp/contests/abc362/submissions/55602146

Try this input:

SpoilerFixed in this submission

thank you loll forgot to print the answer itself

t o y o t a

Edit — solved

how come my O(n^5) code is running for problem e https://atcoder.jp/contests/abc362/submissions/55674291 I think it shouldn't... someone please correct me if i am wrongly calculating complexity or the constraints are just too loose

Can anyone explain me the approach Jinagly used for problem G. https://atcoder.jp/contests/abc362/submissions/55526066 What does the array fail indicate?

it is fail pointer in Aho-Corasick algorithm