#D12757. Fractal Tree

    ID: 10611 Type: Default 2000ms 268MiB

Fractal Tree

Fractal Tree

problem

AOR Ika likes rooted trees with a fractal (self-similar) structure. Consider using the weighted rooted tree T T consisting of N N vertices to represent the rooted tree T T' with the following fractal structure.

  • T T' is the addition of a tree rooted at x x and having a tree structure similar to T T (same cost) for each vertex x x of T T .
  • The root of T T' is the same as that of T T .

The tree represented in this way is, for example, as shown in the figure below.

AOR Ika is trying to do a depth-first search for T T', but finds that it takes a lot of time to trace all the vertices. Therefore, we decided to skip some node visits by performing a depth-first search with a policy of transitioning with a probability of p p and not transitioning with a probability of 1p 1-p at the time of transition during a depth-first search. Given T T and the probability p p , find the expected value of the sum of the costs of all edges to follow when performing a depth-first search for T T'. The information of T T is given by the number of vertices N N and the information of the edges of N1 N-1 , and the vertex 1 1 is the root. Each vertex is labeled 1,2, dots,N 1,2, \ dots, N , and the i (1 lei leN1) i \ (1 \ le i \ le N-1) th edge costs the vertices xi x_i and yi y_i ci c_i It is tied with. The non-deterministic algorithm of the depth-first search that transitions to a child with a probability of p p , which is dealt with in this problem, is expressed as follows. The sum of the costs of the edges that the output  mathrmanswer \ mathrm {answer} follows.

  1. Prepare an empty stack S S .
  2. Set ​​ mathrmanswer=0 ​​\ mathrm {answer} = 0
  3. Push the root vertex of T T' to S S .
  4. Extract the first element of S S and make it x x .
  5. For each child c c of x x , do the following with probability p p and do nothing with probability 1p 1-p .
  • Add vertex c c to S S . Then add the weight of the edge connecting x x to c c to  mathrmanswer \ mathrm {answer} .
  1. If S is not empty, transition to 3.
  2. Output  mathrmanswer \ mathrm {answer} .

output

Print the answer in one line. If the relative or absolute error is less than 106 10 ^ {-6} , then AC. Also, output a line break at the end.

Example

Input

0.75 4 1 2 1 2 3 3 3 4 10

Output

24.8569335938

inputFormat

outputFormat

output  mathrmanswer \ mathrm {answer} follows.

  1. Prepare an empty stack S S .
  2. Set ​​ mathrmanswer=0 ​​\ mathrm {answer} = 0
  3. Push the root vertex of T T' to S S .
  4. Extract the first element of S S and make it x x .
  5. For each child c c of x x , do the following with probability p p and do nothing with probability 1p 1-p .
  • Add vertex c c to S S . Then add the weight of the edge connecting x x to c c to  mathrmanswer \ mathrm {answer} .
  1. If S is not empty, transition to 3.
  2. Output  mathrmanswer \ mathrm {answer} .

output

Print the answer in one line. If the relative or absolute error is less than 106 10 ^ {-6} , then AC. Also, output a line break at the end.

Example

Input

0.75 4 1 2 1 2 3 3 3 4 10

Output

24.8569335938

样例

0.75
4
1 2 1
2 3 3
3 4 10
24.8569335938