Citation Context

.... Therefore, the Landauer erasure cost cancels the extracted work on the average. If the demon wants to operate with maximum efficiency, it must use an optimal coding procedure, i. e., Huffman coding =-=[5]-=-. In this context, the question arises as to how the record length li for the i-th state can be interpreted. Zurek [1] discusses two alternative (sub-optimal) coding procedures for the demon: minimal ...

Citation Context

...iversal ∗ Submitted to IEEE Transactions on Information Theory. † Supported by a fellowship from the Deutsche Forschungsgemeinschaft. 1computer, where the record length is the algorithmic complexity =-=[6]-=- of the state; and Shannon-Fano coding, where the record length is determined by the state’s probability through the inequality − log pi ≤ li < − log pi + 1. The length of a Huffman codeword, on the o...

Citation Context

...rce (N ≥ 2) to which a binary Huffman code is assigned. The i-th letter has probability pi < 1 and codeword length li. The Huffman code can be represented by a binary tree having the sibling property =-=[8]-=- defined as follows: The number of links leading from the root of the tree to a node is called the level of that node. If the level-n node a is connected to the level-(n + 1) nodes b and c, then a is ...

Citation Context

...(m−2)+2 cisely, P − m = where c ≈ 2.27. Introduction j∈{i|li<− log pi−m} pj < 2 −m and P + m = j∈{i|li>−log pi+m} Concepts from information theory gained new importance in physics [1, 2] when Bennett =-=[3]-=- realized that Landauer’s principle [4], which specifies the unavoidable energy cost kBT ln 2 for the erasure of a bit of information, is the clue to the solution of the problem posed by Maxwell’s dem...

Citation Context

...determined by the state’s complexity nor by its probability. Given pi, the Huffman codeword length can, in principle, be as small as 1 bit and as large as [log(( √ 5 + 1)/2)] −1 ≈ 1.44 times − log pi =-=[7]-=-. In this correspondence, we show that the lengths of both Huffman and Shannon-Fano codewords have a similar interpretation. The probability of the states for which the Huffman codeword length differs...

Citation Context

...i. More prepj < 2 −c(m−2)+2 cisely, P − m = where c ≈ 2.27. Introduction j∈{i|li<− log pi−m} pj < 2 −m and P + m = j∈{i|li>−log pi+m} Concepts from information theory gained new importance in physics =-=[1, 2]-=- when Bennett [3] realized that Landauer’s principle [4], which specifies the unavoidable energy cost kBT ln 2 for the erasure of a bit of information, is the clue to the solution of the problem posed...

Citation Context

...i. More prepj < 2 −c(m−2)+2 cisely, P − m = where c ≈ 2.27. Introduction j∈{i|li<− log pi−m} pj < 2 −m and P + m = j∈{i|li>−log pi+m} Concepts from information theory gained new importance in physics =-=[1, 2]-=- when Bennett [3] realized that Landauer’s principle [4], which specifies the unavoidable energy cost kBT ln 2 for the erasure of a bit of information, is the clue to the solution of the problem posed...