I'm not quite sure how to determine what the longest possible codeword is under Huffman encoding for a specific set of frequencies?
Any ideas?
I'm not quite sure how to determine what the longest possible codeword is under Huffman encoding for a specific set of frequencies?
Any ideas?
For a specific set of frequencies? Generate your tree, then see how tall it is... If you're talking about in general, then it's something like N - 1.
Abu-Mostafa and McEliece provide an answer in terms of the probability of the least frequent symbol. The paper also has references to related work on similar questions.