Блог пользователя spookywooky

Автор spookywooky, история, 6 лет назад, По-английски

Working with fractions usually we get a hint like "You should compute P⋅Q−1 modulo 109+7, where Q−1 denotes the multiplicative inverse of Q modulo 109+7."

So, I more (or less) understood what that means. I can express, say 1/5 by the number 400000003. I can calculate that number using https://www.geeksforgeeks.org/fermats-little-theorem/ implemented by some code I found somewhere (see below).

BUT: How do I add (and/or multiply) fractions with huge values?

i.e. how to calculate and express something like this: Let E=10e9+7 Then, how to express: ((E+1) / (E+2)) + ((E+3) / (E+4))

Any hint or link to an understandable explenation would be really helpfull. Thanks.

The code I use so far, based on that fermat thing: ' class Inv { companion object { val defaultMod = 1000000007L var MOD = defaultMod

fun toPower(a: Long, p: Long, mod: Long = MOD): Long {
            var a = a
            var p = p
            var res = 1L
            while (p != 0L) {
                if (p and 1 == 1L)
                    res = res * a % mod
                p = p shr 1
                a = a * a % mod
            }
            return res
        }

        fun inv(x: Long, mod: Long = MOD): Long {
            return toPower(x, mod - 2)
        }

        fun simpleInf(nenner: Long, zaehler: Long): Long {
            return nenner * Inv.inv(zaehler) % Inv.MOD
        }
    }
}

'

  • Проголосовать: нравится
  • +14
  • Проголосовать: не нравится

»
6 лет назад, # |
  Проголосовать: нравится +27 Проголосовать: не нравится

It's pretty simple. Want to add (1/5) mod M and (1/6) mod M? Compute their values (if M = 1e9+7: 400000003, 166666668), add them together, and then mod.

It's the same for multipying.

  • »
    »
    6 лет назад, # ^ |
      Проголосовать: нравится +19 Проголосовать: не нравится

    Thanks, that works fine!

    Further more I found that at subtraction, I need to add M to the result if less than 0, then applying mod if bigger than M.

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

    But what if we want to compare the maximum of two fractions, can we do it by storing PQ - 1modN.

    Addition is pretty clear, but in some DP questions we may also want to store maximum from two previous states where previous states contatins probability, can we store in this form and still do this comparision?

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

      It will overflow if they get way too big, and you can't store modded versions of them, so no. I haven't seen a single DP question requiring this. Maybe they can be solved by greedy or another algorithm, give me an example.

    • »
      »
      »
      5 лет назад, # ^ |
      Rev. 2   Проголосовать: нравится +1 Проголосовать: не нравится

      If previous states store probabilities, then you shouldn't be storing them under a modulo anyways, and instead use doubles. So, atleast this case shouldn't occur in practice.

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

        What do you mean by using doubles?

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

          Sorry if it was unclear. By "double"(s) I meant the data type "double", which stores real numbers as compared to the usual "int" or "long long int" data type which store integers.

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

            The problem I'm facing is, I need to store probabilities of previous states in dp to calculate current state and in the end output probability in the form P.Q^-1 %mod where P and Q are co prime.

            In such a scenario what would be the best to store in a dp? The numerator of the number ?(in this case how to make P and Q coprime?) or P.Q^-1 %mod of the current state?

            • »
              »
              »
              »
              »
              »
              »
              5 лет назад, # ^ |
              Rev. 2   Проголосовать: нравится +1 Проголосовать: не нравится

              Since they require $$$P/Q (mod M)$$$, then you can't use doubles, because of precision error ( modulo enforces you to be exact with the calculations ).

              In general, in these questions, you should look at probability like this $$$Probability( Event A ) = \dfrac{\text{# times A occurs}}{\text{Total number of scenarios}}$$$. So, your $$$dp$$$ should calculate $$$\text{Number of ways of something happening}$$$, and at the end, divide by total scenarios. ( Note: divide here means multiplying by modular inverse ).

              Now, to calculate $$$P/Q ( mod M )$$$, we need what is known as modular inverse of $$$Q$$$. Modular Inverse of $$$Q$$$ under modulo $$$M$$$ ( say $$$X$$$ ), is such a number that $$$ X*Q ( mod M ) = 1 $$$. Read more here.

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

                In that case how can we ensure that P and Q are coprime? According to you number of ways can be stored in the dp and progressively we can find the numerator P using dp. And Q will be total number of scenarios which can be calculated.

                But by finding P and Q seperately how can we ensure they're coprime because 2*modinv(4)!=1*modinv(2).

                I've been trying a question and I feel I'm getting a wrong answer due to this. I've tried everything at this point.

                • »
                  »
                  »
                  »
                  »
                  »
                  »
                  »
                  »
                  5 лет назад, # ^ |
                  Rev. 3   Проголосовать: нравится 0 Проголосовать: не нравится

                  Well, you don't need to make $$$P$$$ and $$$Q$$$ coprime.

                  Proof: Let $$$P$$$ and $$$Q$$$ be not coprime, i.e. they have common factor, say $$$R$$$. Then $$$P = R*A$$$, $$$Q = R*B$$$. Now, $$$inv(Q) = inv(R)*inv(B)$$$

                  So, $$$P*inv(Q) = R*A*inv(R)*B = (R*inv(R))*A*inv(B) = (1)*A*inv(B) $$$. [ $$$inv(R)$$$ is modular inverse of $$$R$$$ ].

                  Also, $$$ 2 * modinv(4) = 1 * modinv(2) $$$. Because $$$ 4 * modinv(4) = 1 $$$, multiply $$$modinv(2)$$$ on both sides, that gives $$$ 2 * modinv(4) = modinv(2) $$$.

                  NOTE: All multiplications are assumed to be computed under the modulo $$$M$$$. I forgot to write it everywhere explicitly.

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

                  Oh yea that makes sense, thanks!

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

        Could you help with how to evaluate p q-1 modulo M M is prime. Given that q = x^y and y can be as large as 10^6

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

    Thanks. It helped :)

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

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

»
5 лет назад, # |
Rev. 2   Проголосовать: нравится +1 Проголосовать: не нравится

I'm very confused about "if q==0,why I should print 0"

Since 0 have no inverses

»
4 года назад, # |
  Проголосовать: нравится +43 Проголосовать: не нравится

I've wrote up a description of how this works. Here it is. Perhaps it will be useful.