Блог пользователя cgy4ever

Автор cgy4ever, история, 7 лет назад, По-английски

Topcoder Open is back! The first round will be 1A:

Time: 12:00 noon EDT Saturday, April 1, 2017

As usual top 250 active coder will advance to Round 2 directly. We will announce the list of coder a bit later, at this time you can check this blog: https://www.topcoder.com/blog/next-algorithm-track-automatic-berths/

You can find rules this year with the schedule here: https://tco17.topcoder.com/algorithm/rules/

Update: Byes are announced here (cut-off rating is 1783): https://tco17.topcoder.com/algorithm/byes/

  • Проголосовать: нравится
  • +80
  • Проголосовать: не нравится

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

cgy4ever I think top 750 will advance to R2 (not 250)

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    It is not what he said. Topcoder's active top 250 advances to Round 2 without taking part in Round 1. Then, for every sub-round of Round 1, top 750 advances to Round 2, as you said.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Contest will start in 23 hours.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Auto comment: topic has been updated by cgy4ever (previous revision, new revision, compare).

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

This round has second Data Science Weekly Challenge associated with it: the author of the fastest submission to Easy problem in Round 1A done in the language which has the least submissions done in it will get a TCO'17 t-shirt!

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится +13 Проголосовать: не нравится

    If by fastest time you mean counting from the time the problem is opened, this seems like a bad challenge as it is trivially cheatable. If this is indeed the case I predict somebody submits Visual Basic.NET solution within 20 seconds of "opening" problem. If it is timed from start of contest ignore my criticism, then it is fine.

    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится +2 Проголосовать: не нравится

      From the start of the contest, of course :-) It also has to be a correct submission, which I obviously forgot to state explicitly.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Can we filer submission by language on Topcoder?

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Will there be a parallel round for those with byes to compete?

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится +8 Проголосовать: не нравится

    No we don't have parallel for round 1. (If we have one then it will be something like Div1 contestants solving Div2 tasks, which is not that fun)

    We will have parallel round for round 2 and 3 for people already advance.

    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится +69 Проголосовать: не нравится

      That is really fun for me! I like solving easy tasks, it gives me a false sense of confidence! Otherwise I will be frustrated by always having difficult problems to work on.

      I wish that all problems on CodeForces and TopCoder were easier so we wouldn't have to think anymore!

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

A bit off topic, and maybe a newbie question, but why can't I practice from past SRMs right now? There's only TCO rounds in the options for practice rooms.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Off topic: does anyone know what to do if arena looks like this on my laptop?

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Are we allowed to compete in TCO 1B if we qualify for the next round today in 1A?

»
7 лет назад, # |
  Проголосовать: нравится +8 Проголосовать: не нравится

987 registrants! I doubt that there will be 750 non-zero scores!

»
7 лет назад, # |
Rev. 2   Проголосовать: нравится 0 Проголосовать: не нравится

I couldn't hack medium problem because I couldn't slide the window.

»
7 лет назад, # |
  Проголосовать: нравится +3 Проголосовать: не нравится

How to solve Hard ?

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    I couldn't code it during contest, but I think it might work. Take the mirror image of the left half of the convex polygon, so both halves are in the right half(x >= 0). Now, you need to find the upper hull of this set of segments. Once you do that, you can sum the volumes by using the volume of a frustum of a cone formula.

    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится 0 Проголосовать: не нравится

      Cool! Thanks :)

    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится 0 Проголосовать: не нравится

      How to find the upper hull ?

      • »
        »
        »
        »
        7 лет назад, # ^ |
        Rev. 3   Проголосовать: нравится +3 Проголосовать: не нравится

        So now you have two sets of segments. Note each set has O(N) segments. We want to find all points of interest, basically, the points which can be present on our upper hull. For that, create a set of points S. Add each vertex belonging to at least one of the segments to S. Now for all segments, if they intersect, add their intersection points to S as well.

        Now create another set Y from S such that Y stores the y coordinate of every point in S. Now our target is for all , find the maximum x that it stretches to. To do that, let's fix the y coordinate, say it is r. Now iterate over all segments, and if the segment is such that it intersects the line y = r, then evaluate the line (assume the segment is a line) by fixing the y coordinate in that line as r, and finding the corresponding x coordinate. Find the maximum such x coordinate for every y.

        Now iterate over all in increasing/decreasing order, and use the formula for volume of a frustum of a cone. The radii for the frustum will be the maximum value of x coordinate for the current y value, and the maximum value of x coordinate for the previous y value. Height will be the difference in the current y coordinate and the previous one.

        AC Code: ideone

        Edit: When you are iterating over all points P, don't add (x, P.y) for every line L. Rather add it for only that line L' such that the corresponding x is maximum

        Edit 2: Rewrote Explanation.

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    Another way of doing it is calculating if only the lefthand side is there, adding if only the righthand side is there, then subtracting the intersection of the two polygons. (Which should be easy to implement if you already have polygon intersection)

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

How to solve Medium?

  • »
    »
    7 лет назад, # ^ |
    Rev. 2   Проголосовать: нравится 0 Проголосовать: не нравится

    dp: f[a][b][c]=maximum surface of a cube of axbxc, in the transition cut a slice of length s from each dimension.

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится +10 Проголосовать: не нравится

    There is another solution to this. I divided each side into parts like A=(s,s,s,s,s....,s+a%s) similarly B and C. Then for every possible combination of a[],b[],c[] I calculated required answer.

    Code
    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится 0 Проголосовать: не нравится

      Optimized version of your solution described in editorial without proving:

      int[] dim = new int[] {A % S + S, B % S + S, C % S + S}; Arrays.sort(dim); dim[0] -= S; return (A * B * C — dim[0] * dim[1] * dim[2]) / S;

      • »
        »
        »
        »
        7 лет назад, # ^ |
          Проголосовать: нравится 0 Проголосовать: не нравится

        This is a greedy approach.

        Let's cut off slices of thickness exactly S (it doesn't make sense to cut off thicker slices). If the total volume of the slices cut is the same, their total area is also the same regardless of the order in which the cuts are done. We can keep cutting them off until a piece with all dimensions less than 2S is left. The dimensions of this piece are given as dim.

        int[] dim = new int[] {A % S + S, B % S + S, C % S + S};
        

        This last piece is also a good slice, with thickness equal to the smallest of its dimensions and area equal to the product of two other dimensions. To make it a slice of thickness S, we can make a cut in the smallest dimension (this preserves the area of the slice). To get the dimensions of the throwaway piece (from which we can't produce any more slices), we model this cut:

        Arrays.sort(dim);
        dim[0] -= S;
        

        Finally, the only part of the original piece that is not converted to slices of thickness exactly S is this throwaway piece. To find the total area of good slices, we take the used volume (the volume of original piece minus the volume of throwaway piece), and divide it by S.

        return (A * B * C - dim[0] * dim[1] * dim[2]) / S;
        
        • »
          »
          »
          »
          »
          7 лет назад, # ^ |
            Проголосовать: нравится 0 Проголосовать: не нравится

          It is not complete proving of optimality. Someone can ask:

          Why we can cut off slices of thickness exactly S of whole piece?

          Why we need stop cutting if dimension less than 2S ?

          • »
            »
            »
            »
            »
            »
            7 лет назад, # ^ |
              Проголосовать: нравится 0 Проголосовать: не нравится

            Because if a dimension less than 2S is cut into a and b (a+b<2S), we would at best be having a>=S and b<S so the part with dimension B is wasted in this case(which could have been used in case the other dimension in the original piece was <a+b) so instead we could just retain the piece as it is and do not cut it.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

So you can literally do nothing and rank 715 :)

»
7 лет назад, # |
  Проголосовать: нравится +15 Проголосовать: не нравится

It is only me or topcoder is too slow ?

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

How to upsolve the problems?

Will the contest be rated? If so, when will the rating be updated?

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

So many failed solutions on 250p problem(mine too)? What was the hack?

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

Will Byes also be there for Round 1B ?

»
7 лет назад, # |
  Проголосовать: нравится +3 Проголосовать: не нравится

When I go to https://www.topcoder.com, I end up in an infinite loop of automatic redirects. Anybody else experiencing this?

»
7 лет назад, # |
Rev. 3   Проголосовать: нравится +3 Проголосовать: не нравится

In medium I am getting better results than expected.

Code

In Test 3: [49,92,20,3] I get 45080 and I should get 30045. What is going on? I would appreciate any insight.

UPD: I got it. The global variable bool inici shall be inside the class to reinitialize it for every test. Sorry for the noise.

»
7 лет назад, # |
Rev. 2   Проголосовать: нравится 0 Проголосовать: не нравится

https://pastebin.com/4kaTEfff

Here is my solution for the 2nd problem. It didn't pass in the contest. I don't know why. I tested it in the practise mode and it showed me "Passed". Also I see many other similar solutions with mine that passed. So can you tell me what is wrong with my source or it is wrong with the system ?

Thank you !

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится
    TLE

    Click

    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится 0 Проголосовать: не нравится

      Well why I get TLE ? Isn't my complexity O(10^6) ? Also why in the practise mode it says "Passed" ?

      • »
        »
        »
        »
        7 лет назад, # ^ |
          Проголосовать: нравится 0 Проголосовать: не нравится

        It's complexity is O(n^4). Declare a global cnt variable, increment it before if(ret != -1) return ret; you can see that there 2.97*10^8 recursive calls. I also had same solution but couldn't submit because of one typo error but it failed in practice room afterwards.

        Most people used O(n^3) solution. Instead of loop, you can cut two blocks one with size s and other size side-s. How to prove this is optimal?

        And there is O(1) solution.

        • »
          »
          »
          »
          »
          7 лет назад, # ^ |
            Проголосовать: нравится 0 Проголосовать: не нравится

          My solution's complexity is O(n4) and it passed the 100 100 100 1 case in 0.24s.

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    I think that's pretty unfortunate. I tested the case provided by dush1729 during contest and that ran in roughly 1.5s in the arena.

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    It's really unfortunate that your code got TLE. But one correction, your complexity is not O(10^6) it's close to O(3*10^8) in worst case. Which is really tight.

    Besides, in your solve function you are using unnecessary too much if else condition which is also a main reason i think. And more importantly when you solve a problem using recursion it always add a constant factor.

»
7 лет назад, # |
  Проголосовать: нравится 0 Проголосовать: не нравится

How do I know if I qualified? do I get an email or something?

  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится 0 Проголосовать: не нравится

    You will get an email. Everyone got a positive score then you will advance (except cheater or ineligible for other reasons).