Re: Fixpoint for LZH
From: hack (hack_at_watson.ibm.com)
Date: 07/15/04
- Next message: Kedi: "Re: How to minimize the absolute sum in curve fitting?"
- Previous message: David Eppstein: "Re: equality vs. identity in Conway's "On Numbers and Games""
- In reply to: Ioannis: "Re: Fixpoint for LZH"
- Next in thread: Ioannis: "Re: Fixpoint for LZH"
- Reply: Ioannis: "Re: Fixpoint for LZH"
- Messages sorted by: [ date ] [ thread ]
Date: 15 Jul 2004 22:05:05 GMT
In article <1089923101.955081@athnrd02.forthnet.gr>,
Ioannis <morpheus@olympus.mons> wrote:
>dev_null wrote:
>
>>
>> And what is about RLE( 12345) = 12345 or RLE( 4321) = 4321 and so on?
>>
>> Regards
>> Luc
>
>Yes, that's theoretically correct. Any string of data that does not
>contain repeating substrings, will be a "fixed point" of RLE.
You're forgetting that you need to be able to *decode* the thing, so you
have to be able to tell a repetition count from an unrepeated value.
Your original example suggested a strict alternation of count and symbol,
so that RLE(12345) would in fact expand to 1112131415. That's ok -- any
lossless compression method must expand some inputs, and the point of the
compression is that, in a given application domain, the objects of interest
have patterns that are mostly compressible.
Once the need for decodability is is taken into account, it becomes a bit
more difficult to find a realistic example.
Michel.
- Next message: Kedi: "Re: How to minimize the absolute sum in curve fitting?"
- Previous message: David Eppstein: "Re: equality vs. identity in Conway's "On Numbers and Games""
- In reply to: Ioannis: "Re: Fixpoint for LZH"
- Next in thread: Ioannis: "Re: Fixpoint for LZH"
- Reply: Ioannis: "Re: Fixpoint for LZH"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|