On the Structure of the Minimum Average Redundancy Code for Monotone Sources
Parvin Rastegari, Mohammadali Khosravifard, Hamed Narimani, T. Aaron Gulliver
Abstract
Parvin Rastegari, Mohammadali Khosravifard, Hamed Narimani, T. Aaron Gulliver
Abstract
The source coding problem is considered where the only information known about the source is the number of symbols n and the order of the symbol probabilities. In this case, the code which achieves the minimum average redundancy is known as the M code. It is the Huffman code for the average distribution of monotone sources, and thus is a fixed code which does not depend on the symbol probabilities. The M code is a near-optimal code not only in the sense of average redundancy, but also in terms of the redundancy for almost all monotone sources. No simple expression is known for the codeword lengths of the M code for a given alphabet size. In this letter, the structure of this code is investigated, and an estimate is given for the number of codewords of a given length. In particular, the number of codewords of length j +1 for encoding 2n symbols is shown to be almost twice the number of codewords of length j for encoding n symbols.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
The source coding problem is considered where the only information known about the source is the number of symbols n and the order of the symbol probabilities. In this case, the code which achieves the minimum average redundancy is known as the M code. It is the Huffman code for the average distribution of monotone sources, and thus is a fixed code which does not depend on the symbol probabilities. The M code is a near-optimal code not only in the sense of average redundancy, but also in terms of the redundancy for almost all monotone sources. No simple expression is known for the codeword lengths of the M code for a given alphabet size. In this letter, the structure of this code is investigated, and an estimate is given for the number of codewords of a given length. In particular, the number of codewords of length j +1 for encoding 2n symbols is shown to be almost twice the number of codewords of length j for encoding n symbols.
Key concepts: Huffman coding, Code word, Prefix code, Universal code, Constant-weight code, Mathematics, Monotone polygon, Systematic code