Talk:Golomb coding

From Wikipedia, the free encyclopedia

Contents

[edit] The first image

Exactly what is the image with q-1, q and q+1 supposed to show? What is the curved arrow at the top for? The 'r' is placed so it looks like a badly typeset qr.

I'm sorry, but the whole image just seems confusing to me. /87.96.186.152 (talk) 10:23, 11 March 2008 (UTC)

[edit] Question on Unary coding...

On the Unary Coding page it says 5 is represented as 11110 yet in the table on the Golomb coding page it says 5 is 111110. Where is the mistake? Or if Golomb coding does use 111110 as 5 and Unary coding uses 11110 as 5, then was it a mistake to say that Golomb uses Unary?

Eric.frederich 16:59, 9 March 2007 (UTC)

The paper "Optimal source codes for geometrically distributed integer alphabets" by Gallager and Van Voorhis (1975, IEEE Transactions on Information Theory IT-21(2):228-230) is the citation "Managing Gigabytes" gives for the probability distribution quoted in the article, but I haven't had a chance to read the paper myself, so I'm hesitant to add it to the refrences. I'll get to it eventually, but if someone else wants to read it and see if it should be refrenced, please do.Mark Tozzi 06:47, 20 December 2005 (UTC)

their paper is not so great. i can send it to you if you provide email.. []

[edit] Insufficient Context

This article could use a bit more introductory context to orient non-expert readers.

For example:

  • 'entropy' is introduced without context (cryptanalysis? astrophysics? neoclassical economics? what sense of entropy is relevant here?). Which connotation is not necessarily obvious to those unfamiliar.
  • a quick 1 sentence blurb about what this kind of coding is used for would help (eg data compression, natural language processing, lexical analysis, string functions, whatever). It is implied in the text, but intro could be beefed up a bit.

dr.ef.tymac 16:27, 19 November 2006 (UTC)

I agree with the comments above about insufficient context, but I'm also confused about Rice's contribution to Golomb-Rice codes. The introduction says Golomb's scheme (which references a paper dated 1966) generalized Rice's scheme which was developed earlier, except that the only reference to Rice's work is a paper dated 1979. If Rice really did invent something before 1966, where's the reference? Otherwise, why is Rice's name even associated with the codes?

[edit] Question on alphabets

Quote from the article: "Golomb coding can also be used (as in Golomb's original article) to encode an alphabet of two symbols, where one is more likely than the other. In this case it can be thought of as a kind of run-length encoding."

As far as I have understood the concept of the code from reading the original paper of S. Golomb, the code ALWAYS operates via a two-symbols alphabet, as it is used to produce a binary representation of a given number; hence, it is a mapping of that particular number (i.e., 42) to a sequence of the both sybmols occuring in the alphabet (obviously, 0 and 1), and vice versa. The length of the sequence is determined by the probability for one of those symbols to arise (for the case M = 10, this probability is 2^(-1/10) = 0,933 for one of the symbols; it is a question of convention if it would be called 0 or 1). If we were to code a number in a ternary representation, our alphabet would contain a third symbol, i.e. the ternary digit 2.

In the light of the above thoughts, I think that there is a slight confusion in the article in the sense that the number to be coded is regarded as part of the alphabet.

The quote from the article is completely misleading and confusing; it can be interpreted as if the numbers we would like to code were only two; however, this is not the case; it should mean that the base (or alphabet) used to produce some output consists of only two symbols - which is already clear, since we are talking of a binary code.

There should be a clear distinguishment between the set of numbers to be coded (i.e., 1, 2, ..., n), and the alphabet used to produce the code of the numbers in that set.

Nikolay Petkov 7c0[at]mail.com 22:27, 18 May 2007 (EET)