51

Re: About one method data compressions

FXS wrote:

is not present.
If an initial line -

 "10101011100111010101 [b] 1111 [/b] 1001111001 [b] 1111111111 [/b]" 

, That a flag "000", as well as is described in item 1
If the casual line, suddenly, turned out

 "11111111111111111111111111111111111111111111" 

, That oblate will be:

 "10101011100111010101 [b] 000 [/b] 100111001 [b] 000 [/b]" 

(Separately it will be necessary to transfer plus of length of two "" segments.

And where seed for initialization ?

52

Re: About one method data compressions

Dima T;
Too it will be necessary to transfer, it is already written - to item 7 of an initial post.

53

Re: About one method data compressions

Dima T wrote:

for storage seed it is required places as much how many initial substring occupies, it I above proved

I this your proof not , excuse. What for "initial under - a line"?

54

Re: About one method data compressions

FXS wrote:

Dima T;
Too it will be necessary to transfer, it is already written - to item 7 of an initial post.

In a case if substring repetitions it will be found 2 and more finding of the shaken variant becomes more probable, but the size shaken will be less on any units of bits.
And I invented as this task  for one pass of all variants seed smile

55

Re: About one method data compressions

FXS wrote:

it is passed...
I this your proof not , excuse. What for "initial under - a line"?

For example that was 1111 (it is initial substring), i.e. the minimum of 2^4 variants seed should have 4 bits , therefore to write down specific value seed it will be necessary 4 bits.
Though for your algorithm the proof is not absolutely correct, it is necessary since to find any one of all set, i.e. probability above, but I feel that not much more, it is necessary to invent as it to justify.

56

Re: About one method data compressions

Dima T;
"For example that was 1111 (it is initial substring)"... in one specific position ... " should have a minimum of 2^4 variants seed"
-- But to us it is absolutely not mandatory, that was "1111" and in this position.

57

Re: About one method data compressions

FXS wrote:

- but to us it is absolutely not mandatory, that was "1111" and in this position.

Truly, I therefore wrote "the proof not absolutely correctly".
It is necessary to be repelled from another: what . substring most possibly is? Possible combinations seed think M of bits, where M such that 2^M most less. Because  can produce most of all various combinations on M of bits.
Further it is necessary to count somehow probability of obtaining of coincidence M+1 of bits, M+2 etc. I will not think as, but I feel that everyone . reduces probability twice, i.e. to spare pair byte it is already enough rare occurence.

58

Re: About one method data compressions

Siemargl wrote:

It is normal algorithm LZW, only with errors and accidentally-generuemym dictionary:fail:

Only you whether did not note, whether decided to hold back that "the generated dictionary" means that the dictionary not to put in the output file (increasing its size) as it becomes in LZW.
("With errors" I did not understand, therefore I can not concern.)

59

Re: About one method data compressions

FXS wrote:

the dictionary not needs to be put in the output file

60

Re: About one method data compressions

Now, almost in two weeks, I in another way would formulate the compression concept (hypothetical, of course, as its practicability is not proved). Namely, I would replace "flag" (which is not clear-what-length) with "chain addressing":
There is data  - a bit line of length N, - which is required to be compressed, and standard .
1. In a cycle we sort out values seed (s), used for initialization :  (s).
2. We write down s in the output file beginning.
3. By means of  (s) it is generated casual sequence (joint venture) of the same length N.
4. We install "the counter of chain addressing" in a zero: =0.
5. Comparing   and the joint venture, we move along both of them until it is found in them (to "the chain address") identical substring () by length of M.
6. If the notation of record of integer numbers used by us allow to write down these two numbers (And and) is shorter, than M of bits it is writeable them in the output file beginning. Also we write down in its end the next portion of not oblate bits - that went to  (that is substring  "is thrown out", we do not write to the output file).
(6.1. By the way, we can decide that with substrings is shorter 0 to communicate in essence does not follow; it is necessary then to write down not M, and difference -0 that is a little more favourable.)
7. If the condition described in item 6, is fulfilled, we pass to item 4; if it is not fulfilled - we pass to item 5. The further scanning of lines - anyway - proceeds from the first bit after Item substring
8. Reaching line end , we estimate the received variant of compression: if it is the best previous, we save it if is not present - is forgotten. Also we pass to item 1.