Why is my probability so far off with this Binary Tree using?

Question

It's basically just an implementation of the Huffman Coding algorithm, but when I check the probability of the end BinaryTree (the only item in the queue left) it's grossly high.

// Make a BinaryTree for each item in CharOccurrences and add as an entry in initialQueue
for (int i = 0; i < charOccurrences.size(); i++) {
  BinaryTree<CharProfile> bTree = new BinaryTree<CharProfile>();
  bTree.makeRoot(charOccurrences.get(i));
  initialQueue.add(bTree);
}

// Create the BinaryTree that we're adding to the resultQueue
BinaryTree<CharProfile> treeMerge = new BinaryTree<CharProfile>();

// Create the CharProfile that will hold the probability of the two merged trees
CharProfile data;

while (!initialQueue.isEmpty()) {
  // Check if the resultQueue is empty, in which case we only need to look at initialQueue
  if (resultQueue.isEmpty()) {
    treeMerge.setLeft(initialQueue.remove());
    treeMerge.setRight(initialQueue.remove());

    // Set treeMerge's data to be the sum of its two child trees' probabilities with a null char value
    data = new CharProfile('\0');
    data.setProbability(treeMerge.getLeft().getData().getProbability() + treeMerge.getRight().getData().getProbability());
    treeMerge.setData(data);
  }
  else {
    // Set the left part of treeMerge to the lowest of the front of the two queues
    if (initialQueue.peek().getData().getProbability() <= resultQueue.peek().getData().getProbability()) {
      treeMerge.setLeft(initialQueue.remove());
    }
    else {
      treeMerge.setLeft(resultQueue.remove());
    }

    if (!initialQueue.isEmpty()) {
      // Set the right part of treeMerge to the lowest of the front of the two queues
      if (initialQueue.peek().getData().getProbability() <= resultQueue.peek().getData().getProbability()) {
        treeMerge.setRight(initialQueue.remove());
      }
      else {
        treeMerge.setRight(resultQueue.remove());
      }
    }
    // In the case that initialQueue is now empty (as a result of just dequeuing the last element), simply make the right tree resultQueue's head
    else {
      treeMerge.setRight(resultQueue.remove());
    }

    // Set treeMerge's data to be the sum of its two child trees' probabilities with a null char value
    data = new CharProfile('\0');
    data.setProbability(treeMerge.getLeft().getData().getProbability() + treeMerge.getRight().getData().getProbability());
    treeMerge.setData(data);
  }

  // Add the new tree we create to the resultQueue
  resultQueue.add(treeMerge);
}

if (resultQueue.size() > 1) {
  while (resultQueue.size() != 1) {
    treeMerge.setLeft(resultQueue.remove());
    treeMerge.setRight(resultQueue.remove());

    data = new CharProfile('\0');
    data.setProbability(treeMerge.getLeft().getData().getProbability() + treeMerge.getRight().getData().getProbability());
    treeMerge.setData(data);

    resultQueue.add(treeMerge); 
  }
}

I then have this at the end:

System.out.println("\nProbability of end tree: " 
    + resultQueue.peek().getData().getProbability());

Which gives me:

Probability of end tree: 42728.31718061674

Diego · Accepted Answer · 2012-11-04T19:58:01.270

4

Move the following lines inside the while loop:

// Create the BinaryTree that we're adding to the resultQueue
BinaryTree<CharProfile> treeMerge = new BinaryTree<CharProfile>();

Otherwise, one iteration adds treeMerge to resultQueue, and the next one may do treeMerge.setLeft(resultQueue.remove());, which makes treeMerge a child of itself...

edited Nov 04 '12 at 19:58

answered Nov 04 '12 at 18:28

Diego

18,035
5
62
66

Thank you, that was it. I'll award your bounty as soon as it lets me. :) – Doug Smith Nov 04 '12 at 23:49

Why is my probability so far off with this Binary Tree using?

1 Answers1