Science and Technology news... :: View topic - The PERFECT EQUATION = LOSSLESS COMPRESSION!!!! TRUE!!!!

Thomas Richter · Guest

jacko schrieb:
[quote]On Jun 17, 9:15 pm, Thomas Richter <t...@math.tu-berlin.de> wrote:
jacko wrote:
Excuse me, but if you want to be understood, please use proper language.
What is a 50:50 entropy? That term is not defined. Entropy requires a
model (hence, an alphabet of symbols), what is it? What are the
probabilities, what is the size of the alphabet, what are the symbols,
and which entropy do you mean?
Yes Sum -p log2(p)
so I would say 50:50 % implies 1/2 0 1/2 1 entropy =1.(as useual most
things like mark space ratio written in 100% total)
So you have a binary alphabet with two symbols of a memoryless source.
Why is this a good model of the "carrier"? Or rather, why would it make
sense to understand this as "random source"? For example, the sequence

A never said it was a memoryless source, in fact it has to have some
memory in order to have a lower entropy than 1.
[/quote]
Huh? If you compute the entropy like you did (zero-order entropy), then
of course it can have a small entropy without a memory. Consider a dice-roll
where you map 1->0 and all other events to one. The corresponding random
source is of course memoryless, but has a small entropy.

[quote]I never said the
source was random, mearly pseudo-random, I don>t think you will get
far reversing any truly random carrier source.
010101010101010101010101010101010101....

can be easily generated "algorithmically" and is even a "typical"
sequence for an i.i.d. zero-order maximum entropy two-symbol random
source. Is that a useful carrier? If so, why not pick it always (and
simplify the design), if not so, why not? Why is this a useful
description of a "carrier"?

You will note the source entropy of this "unideal" carrier excedds the
required entropy, and so is very unuseful as a carrier.
[/quote]
Why is it "unideal"? What qualifies an "ideal" carrier?

[quote]I think what is a bit bizarre here is that you state that your encoder
is in full control of generating the *carrier*, so it is not a random
source, so why is it necessary or useful to model it as such.

You could treat it as a totally ordered system, and define the source
entropy as the entropy of the sequence which is produced. It is useful
to model it as such, due to the nature of calculating entropy.
[/quote]
So I>m now confused. The entropy you compute is the entropy of a zero
order memoryless source (which might be an applicable model for a pseudo-
random generator, one can choose to take this view, why not), still you
say the source requires memory?

[quote]Specifically, given a method I would expect the outcome to depend
heavily on *which* "carrier" you pick, not only on its zero-order iid
random source property.

The carrier picked would definatly be the most important thing to
consider. If the carrier produces at least the two 0s required later
in the order of the sequence inorder to allow recollection of all
emission done for each 1 in the carrier, not much else would matter.
one of these 0s will be modulated to a 1, and the other will be left
as a 0. It the main sense, it does not matter what actual sequence is
traversed, just that traversal leads to a net collection from the
input stream which modulates the carrier.
[/quote]
So given a source sequence, how to pick the proper carrier? And how to let
the receiver know that you picked a specific carrier (you would need to
inform the receiver of course on your choice, otherwise you cannot reverse
the process). And assume that I have an algorithm to pick the proper carrier,
is it clear that for every source a "suitable" carrier exist? Or that it can
be picked from a small family? How about the size of the side-information
required to let the receiver know about your choice? Is that included?

This said, I might be convinced that for every source a "carrier" exists
that "modulates" this source to complete zero, hence something very
compressible. However, this gains you nothing since then the "carrier" is so
complicated (probably more complicated than the source) and you would need
to transmit this "side information", so the net effect is no compression
anymore.

IOW, why is this a good idea? A quick argument I have would go as follows:
If the "carrier" absorbs entropy, this entropy doesn>t go to nowhere. It is
exactly the amount of information required to identify the carrier successfully
at the receiver.

[quote]if(B==1) emit(A,1); //emit rather than absorb
if(B==0) if(collect(A)==1) { B=1;emit(A,0); }
What does "emit(A,1)" mean? Generate two output symbols, first A,then 1?
What does "collect(A)==1" mean? Why don>t you need to emit the input
signal (I suppose "A") in case "B == 0 and collect(A) != 1"?
Why would you expect the output of this be useful?
What is the carrier, what is the "source"? (I suppose A is the source,
but you never state this.)

A is the user data stream
emit(A,x) = "emit the binary symbol x to the beginning of straem A"
collect(A) = "collect binary symbol from the beginning of stream A"
[/quote]
I don>t understand - how do you emit a symbol at the beginning of a stream?
Do you mean, your input is a FIFO, not a stream? (A stream is something I would
consider to be read and written only in one direction)

[quote]The carrier is B, The modulation is the B=1; statement using C
language syntax.
[/quote]
Yes, that part I understood. I still don>t understand the algorithm.

[quote]I don>t quite follow you - are you considering a multi-channel coding
setup? What is the "baseband" (I don>t see a definition). The "modulated
carrier" is likely the output of the above pseudo-code (which seems
incomplete.). If I count correctly, the above should expand the source
by a factor of two. Why is this a good thing to do?

if B =
000000000000000100000000001000000010000000000000000000000000010000000000000000011000000000000
and pick user data A as any bit sequence you want.

just use that one first so you can see the advantage of a low entropy
carrier.
baseband = in TV it would be the 8MHz UK PAL I signal from DC to 8MHz,
and so by extension I would say it was the thing that will be
modulated on to a carrier.
[/quote]
In your above algorithm, B seems to indicate a state of a state machine - I never
see a statement that B is "collected" from the B-stream. Is that an omission?
Or is B updated at the end of each algorithm step? If so, why does line 2 set B to one?

[quote]I am not realy interested in the carrier output as a bit by bit stream
to send anywhere. I am interested in the register value storing the
current total state of the carrier generator. So that I may instance a
copy of the carrier generator elsewhere and rewind it.!!!!!
[/quote]
Now I>m totally confused. Up to now I had the impression that the carrier is an
input to your encoder, not an output. If it is an output, why is there no "emit(B,x)"
in the above?

[quote]Why would you believe that the above pseudo-code causes an entropy
reduction? Or at least a reduction that compensates for the factor of
two of the expansion generated by the algorithm?

The above code does not necessarily cause an entropy reduction. The
above code allows modulation of any entropy stream onto a carrier
which has already had its entropy reduced by algorithmic design.
[/quote]
Ok, so why do that then? IOW, how would you require the carrier to look like to
get an output of lower entropy?

[quote]sending the current seed of B is like sending B and all of A collected
upon it.
[/quote]
Well - if B is generated by a Pseudo-random generator, I agree. But the amount
of size reduction should be no larger than the size of the seed, I would believe.

Or, even different, why would I expect to find for a given source a good pseudo-random
generator and a good seed that would help me? This seems rather hard.

So long,
Thomas

jacko · Guest

Thomas Richter wrote:
[quote]jacko schrieb:
On Jun 17, 9:15 pm, Thomas Richter <t...@math.tu-berlin.de> wrote:
jacko wrote:
Excuse me, but if you want to be understood, please use proper language.
What is a 50:50 entropy? That term is not defined. Entropy requires a
model (hence, an alphabet of symbols), what is it? What are the
probabilities, what is the size of the alphabet, what are the symbols,
and which entropy do you mean?
Yes Sum -p log2(p)
so I would say 50:50 % implies 1/2 0 1/2 1 entropy =1.(as useual most
things like mark space ratio written in 100% total)
So you have a binary alphabet with two symbols of a memoryless source.
Why is this a good model of the "carrier"? Or rather, why would it make
sense to understand this as "random source"? For example, the sequence

A never said it was a memoryless source, in fact it has to have some
memory in order to have a lower entropy than 1.

Huh? If you compute the entropy like you did (zero-order entropy), then
of course it can have a small entropy without a memory. Consider a dice-roll
where you map 1->0 and all other events to one. The corresponding random
source is of course memoryless, but has a small entropy.

bad carrier example due to not being able to hold modulation of 1 to 0[/quote]
as bias favours 1. and so which 1 state is the correct 1 state to
demodulate to 1. well a carrier has to hold modulation or it has not
been modulated in the strictest sense.

[quote]I never said the
source was random, mearly pseudo-random, I don>t think you will get
far reversing any truly random carrier source.
010101010101010101010101010101010101....

can be easily generated "algorithmically" and is even a "typical"
sequence for an i.i.d. zero-order maximum entropy two-symbol random
source. Is that a useful carrier? If so, why not pick it always (and
simplify the design), if not so, why not? Why is this a useful
description of a "carrier"?

You will note the source entropy of this "unideal" carrier excedds the
required entropy, and so is very unuseful as a carrier.

Why is it "unideal"? What qualifies an "ideal" carrier?
one which holds modulation.
I think what is a bit bizarre here is that you state that your encoder
is in full control of generating the *carrier*, so it is not a random
source, so why is it necessary or useful to model it as such.

You could treat it as a totally ordered system, and define the source
entropy as the entropy of the sequence which is produced. It is useful
to model it as such, due to the nature of calculating entropy.

So I>m now confused. The entropy you compute is the entropy of a zero
order memoryless source (which might be an applicable model for a pseudo-
random generator, one can choose to take this view, why not), still you
say the source requires memory?
if it has no memory, how are you going to make it modulatable?
Specifically, given a method I would expect the outcome to depend
heavily on *which* "carrier" you pick, not only on its zero-order iid
random source property.

The carrier picked would definatly be the most important thing to
consider. If the carrier produces at least the two 0s required later
in the order of the sequence inorder to allow recollection of all
emission done for each 1 in the carrier, not much else would matter.
one of these 0s will be modulated to a 1, and the other will be left
as a 0. It the main sense, it does not matter what actual sequence is
traversed, just that traversal leads to a net collection from the
input stream which modulates the carrier.

So given a source sequence, how to pick the proper carrier? And how to let
the receiver know that you picked a specific carrier (you would need to
inform the receiver of course on your choice, otherwise you cannot reverse
the process). And assume that I have an algorithm to pick the proper carrier,
is it clear that for every source a "suitable" carrier exist? Or that it can
be picked from a small family? How about the size of the side-information
required to let the receiver know about your choice? Is that included?
carrier does not depend on source.
This said, I might be convinced that for every source a "carrier" exists
that "modulates" this source to complete zero, hence something very
compressible. However, this gains you nothing since then the "carrier" is so
complicated (probably more complicated than the source) and you would need
to transmit this "side information", so the net effect is no compression
anymore.
file is collected and so zero size.
IOW, why is this a good idea? A quick argument I have would go as follows:
If the "carrier" absorbs entropy, this entropy doesn>t go to nowhere. It is
exactly the amount of information required to identify the carrier successfully
at the receiver.

are you sure. are you sure it has to be added somewhere, and may not[/quote]
be subtracted from somewhere?
[quote]
if(B==1) emit(A,1); //emit rather than absorb
if(B==0) if(collect(A)==1) { B=1;emit(A,0); }
What does "emit(A,1)" mean? Generate two output symbols, first A,then 1?
What does "collect(A)==1" mean? Why don>t you need to emit the input
signal (I suppose "A") in case "B == 0 and collect(A) != 1"?
Why would you expect the output of this be useful?
What is the carrier, what is the "source"? (I suppose A is the source,
but you never state this.)

A is the user data stream
emit(A,x) = "emit the binary symbol x to the beginning of straem A"
collect(A) = "collect binary symbol from the beginning of stream A"

I don>t understand - how do you emit a symbol at the beginning of a stream?
Do you mean, your input is a FIFO, not a stream? (A stream is something I would
consider to be read and written only in one direction)
fifo if easier to understand.
The carrier is B, The modulation is the B=1; statement using C
language syntax.

Yes, that part I understood. I still don>t understand the algorithm.
do you understand the modulation technique, otherwise what>s the point[/quote]
in describing a carrier to you.
[quote]I don>t quite follow you - are you considering a multi-channel coding
setup? What is the "baseband" (I don>t see a definition). The "modulated
carrier" is likely the output of the above pseudo-code (which seems
incomplete.). If I count correctly, the above should expand the source
by a factor of two. Why is this a good thing to do?

if B =
000000000000000100000000001000000010000000000000000000000000010000000000000000011000000000000
and pick user data A as any bit sequence you want.

just use that one first so you can see the advantage of a low entropy
carrier.
baseband = in TV it would be the 8MHz UK PAL I signal from DC to 8MHz,
and so by extension I would say it was the thing that will be
modulated on to a carrier.

In your above algorithm, B seems to indicate a state of a state machine - I never
see a statement that B is "collected" from the B-stream. Is that an omission?
Or is B updated at the end of each algorithm step? If so, why does line 2 set B to one?
b is not collected but is evaluated once per cycle.
I am not realy interested in the carrier output as a bit by bit stream
to send anywhere. I am interested in the register value storing the
current total state of the carrier generator. So that I may instance a
copy of the carrier generator elsewhere and rewind it.!!!!!

Now I>m totally confused. Up to now I had the impression that the carrier is an
input to your encoder, not an output. If it is an output, why is there no "emit(B,x)"
in the above?
it is a state of a state machine, so sending it somewhere allows[/quote]
instancing the state machine.
[quote]Why would you believe that the above pseudo-code causes an entropy
reduction? Or at least a reduction that compensates for the factor of
two of the expansion generated by the algorithm?

The above code does not necessarily cause an entropy reduction. The
above code allows modulation of any entropy stream onto a carrier
which has already had its entropy reduced by algorithmic design.

Ok, so why do that then? IOW, how would you require the carrier to look like to
get an output of lower entropy?
the carrier has by design (just one design) a low entropy already we[/quote]
are not trying to use modulation to lower it, but to send information
on it.
[quote]sending the current seed of B is like sending B and all of A collected
upon it.

Well - if B is generated by a Pseudo-random generator, I agree. But the amount
of size reduction should be no larger than the size of the seed, I would believe.
why would that be? the collection count is based only on the carrier[/quote]
entrpy.
[quote]Or, even different, why would I expect to find for a given source a good pseudo-random
generator and a good seed that would help me? This seems rather hard.

So long,
Thomas
[/quote]
windows is shit again.

jacko · Guest

Thomas Richter wrote:
[quote]jacko schrieb:

So you have a binary alphabet with two symbols of a memoryless source.
Why is this a good model of the "carrier"? Or rather, why would it make
sense to understand this as "random source"? For example, the sequence
A never said it was a memoryless source, in fact it has to have some
memory in order to have a lower entropy than 1.
Huh? If you compute the entropy like you did (zero-order entropy), then
of course it can have a small entropy without a memory. Consider a dice-roll
where you map 1->0 and all other events to one. The corresponding random
source is of course memoryless, but has a small entropy.

bad carrier example due to not being able to hold modulation of 1 to 0
as bias favours 1. and so which 1 state is the correct 1 state to
demodulate to 1. well a carrier has to hold modulation or it has not
been modulated in the strictest sense.

Why is this an answer to my question? Did you get my question?
the entropy may be different but only the computed 0th order has value[/quote]
at present for calculating the nett collection of A.
[quote]You will note the source entropy of this "unideal" carrier excedds the
required entropy, and so is very unuseful as a carrier.
Why is it "unideal"? What qualifies an "ideal" carrier?
one which holds modulation.

What does that mean? How can "modulation hold"?
if you modulated a 1 to 0 (major to minor state) how would you know in[/quote]
the dice example which major state to demodulate back to? it does
therefore not hold modulation.
[quote]So given a source sequence, how to pick the proper carrier? And how to let
the receiver know that you picked a specific carrier (you would need to
inform the receiver of course on your choice, otherwise you cannot reverse
the process). And assume that I have an algorithm to pick the proper carrier,
is it clear that for every source a "suitable" carrier exist? Or that it can
be picked from a small family? How about the size of the side-information
required to let the receiver know about your choice? Is that included?
carrier does not depend on source.

If it doesn>t, why would you expect that "modulation" helps?
if the file size of A is going toward zero, I say that helps. when it[/quote]
becomes 0 it has helped a lot.
[quote]This said, I might be convinced that for every source a "carrier" exists
that "modulates" this source to complete zero, hence something very
compressible. However, this gains you nothing since then the "carrier" is so
complicated (probably more complicated than the source) and you would need
to transmit this "side information", so the net effect is no compression
anymore.
file is collected and so zero size.
IOW, why is this a good idea? A quick argument I have would go as follows:
If the "carrier" absorbs entropy, this entropy doesn>t go to nowhere. It is
exactly the amount of information required to identify the carrier successfully
at the receiver.

are you sure. are you sure it has to be added somewhere, and may not
be subtracted from somewhere?

Did you get the question? I did not mean "added" in the algebraic sense,
but rather, it (or the seed) has to be included in the transmission to
make it decodeable. However, as the carrier becomes more and more
complicated, more bits have to be spend to transmit the "seed".
it does not have to be so. if it is a self lowering entropy stream. a[/quote]
special carrier.
[quote]if(B==1) emit(A,1); //emit rather than absorb
if(B==0) if(collect(A)==1) { B=1;emit(A,0); }
What does "emit(A,1)" mean? Generate two output symbols, first A,then 1?
What does "collect(A)==1" mean? Why don>t you need to emit the input
signal (I suppose "A") in case "B == 0 and collect(A) != 1"?
Why would you expect the output of this be useful?
What is the carrier, what is the "source"? (I suppose A is the source,
but you never state this.)
A is the user data stream
emit(A,x) = "emit the binary symbol x to the beginning of straem A"
collect(A) = "collect binary symbol from the beginning of stream A"
I don>t understand - how do you emit a symbol at the beginning of a stream?
Do you mean, your input is a FIFO, not a stream? (A stream is something I would
consider to be read and written only in one direction)
fifo if easier to understand.

So the input is not a stream, but a FIFO.
if it helps to understand then yes. a fifo with a stream preloaded[/quote]
into it. the user content.
[quote]The carrier is B, The modulation is the B=1; statement using C
language syntax.
Yes, that part I understood. I still don>t understand the algorithm.
do you understand the modulation technique, otherwise what>s the point
in describing a carrier to you.

Look, "understanding the modulation technique" and "understanding the
algorithm" mean exactly the same. I cannot understand one without the other.
no they don>t understanding the modulation onto a special carrier and[/quote]
the special carrier design are different.
[quote]In your above algorithm, B seems to indicate a state of a state machine - I never
see a statement that B is "collected" from the B-stream. Is that an omission?
Or is B updated at the end of each algorithm step? If so, why does line 2 set B to one?
b is not collected but is evaluated once per cycle.

Where is that "evaluation" step in the algorithm above? I.e. when is B
updated? There is no statement that changes the value of B in the above
algorithm.
b=1 or did you not read it. the evaluation is per carrier cycle.
The above code does not necessarily cause an entropy reduction. The
above code allows modulation of any entropy stream onto a carrier
which has already had its entropy reduced by algorithmic design.
Ok, so why do that then? IOW, how would you require the carrier to look like to
get an output of lower entropy?
the carrier has by design (just one design) a low entropy already we
are not trying to use modulation to lower it, but to send information
on it.

Ok, but then what>s the point if the point is not to lower entropy? IOW,
what does "absorb" then mean, and why do you want to do that?
to get any stream collected/absorbed onto a carrier of much lower[/quote]
complexity.
[quote]IOW, what>s the purpose of this algorithm? Don>t tell me "modulation",
that>s just a different word. Just tell me "why would I want to run it".
if you had a special carrier you would run it to collect fifo A onto[/quote]
it to compress data.
[quote]sending the current seed of B is like sending B and all of A collected
upon it.
Well - if B is generated by a Pseudo-random generator, I agree. But the amount
of size reduction should be no larger than the size of the seed, I would believe.
why would that be? the collection count is based only on the carrier
entrpy.

No, the collection count is based on the number of "1"s in the B-stream.
That>s not the entropy.
is based on != is.
Or, even different, why would I expect to find for a given source a good pseudo-random
generator and a good seed that would help me? This seems rather hard.

windows is shit again.

How does that relate to anything I>ve written?
it doesn>t it makes my writing shorter via pocket pc not desktop which[/quote]
has said 8 minuites left for the last 2 hrs.
[quote]
So long,
Thomas[/quote]

Thomas Richter · Guest

jacko schrieb:

[quote]So you have a binary alphabet with two symbols of a memoryless source.
Why is this a good model of the "carrier"? Or rather, why would it make
sense to understand this as "random source"? For example, the sequence
A never said it was a memoryless source, in fact it has to have some
memory in order to have a lower entropy than 1.
Huh? If you compute the entropy like you did (zero-order entropy), then
of course it can have a small entropy without a memory. Consider a dice-roll
where you map 1->0 and all other events to one. The corresponding random
source is of course memoryless, but has a small entropy.

bad carrier example due to not being able to hold modulation of 1 to 0
as bias favours 1. and so which 1 state is the correct 1 state to
demodulate to 1. well a carrier has to hold modulation or it has not
been modulated in the strictest sense.
[/quote]
Why is this an answer to my question? Did you get my question?

[quote]You will note the source entropy of this "unideal" carrier excedds the
required entropy, and so is very unuseful as a carrier.
Why is it "unideal"? What qualifies an "ideal" carrier?
one which holds modulation.
[/quote]
What does that mean? How can "modulation hold"?

[quote]So given a source sequence, how to pick the proper carrier? And how to let
the receiver know that you picked a specific carrier (you would need to
inform the receiver of course on your choice, otherwise you cannot reverse
the process). And assume that I have an algorithm to pick the proper carrier,
is it clear that for every source a "suitable" carrier exist? Or that it can
be picked from a small family? How about the size of the side-information
required to let the receiver know about your choice? Is that included?
carrier does not depend on source.
[/quote]
If it doesn>t, why would you expect that "modulation" helps?

[quote]This said, I might be convinced that for every source a "carrier" exists
that "modulates" this source to complete zero, hence something very
compressible. However, this gains you nothing since then the "carrier" is so
complicated (probably more complicated than the source) and you would need
to transmit this "side information", so the net effect is no compression
anymore.
file is collected and so zero size.
IOW, why is this a good idea? A quick argument I have would go as follows:
If the "carrier" absorbs entropy, this entropy doesn>t go to nowhere. It is
exactly the amount of information required to identify the carrier successfully
at the receiver.

are you sure. are you sure it has to be added somewhere, and may not
be subtracted from somewhere?
[/quote]
Did you get the question? I did not mean "added" in the algebraic sense,
but rather, it (or the seed) has to be included in the transmission to
make it decodeable. However, as the carrier becomes more and more
complicated, more bits have to be spend to transmit the "seed".

[quote]if(B==1) emit(A,1); //emit rather than absorb
if(B==0) if(collect(A)==1) { B=1;emit(A,0); }
What does "emit(A,1)" mean? Generate two output symbols, first A,then 1?
What does "collect(A)==1" mean? Why don>t you need to emit the input
signal (I suppose "A") in case "B == 0 and collect(A) != 1"?
Why would you expect the output of this be useful?
What is the carrier, what is the "source"? (I suppose A is the source,
but you never state this.)
A is the user data stream
emit(A,x) = "emit the binary symbol x to the beginning of straem A"
collect(A) = "collect binary symbol from the beginning of stream A"
I don>t understand - how do you emit a symbol at the beginning of a stream?
Do you mean, your input is a FIFO, not a stream? (A stream is something I would
consider to be read and written only in one direction)
fifo if easier to understand.
[/quote]
So the input is not a stream, but a FIFO.

[quote]The carrier is B, The modulation is the B=1; statement using C
language syntax.
Yes, that part I understood. I still don>t understand the algorithm.
do you understand the modulation technique, otherwise what>s the point
in describing a carrier to you.
[/quote]
Look, "understanding the modulation technique" and "understanding the
algorithm" mean exactly the same. I cannot understand one without the other.

[quote]In your above algorithm, B seems to indicate a state of a state machine - I never
see a statement that B is "collected" from the B-stream. Is that an omission?
Or is B updated at the end of each algorithm step? If so, why does line 2 set B to one?
b is not collected but is evaluated once per cycle.
[/quote]
Where is that "evaluation" step in the algorithm above? I.e. when is B
updated? There is no statement that changes the value of B in the above
algorithm.

[quote]The above code does not necessarily cause an entropy reduction. The
above code allows modulation of any entropy stream onto a carrier
which has already had its entropy reduced by algorithmic design.
Ok, so why do that then? IOW, how would you require the carrier to look like to
get an output of lower entropy?
the carrier has by design (just one design) a low entropy already we
are not trying to use modulation to lower it, but to send information
on it.
[/quote]
Ok, but then what>s the point if the point is not to lower entropy? IOW,
what does "absorb" then mean, and why do you want to do that?

IOW, what>s the purpose of this algorithm? Don>t tell me "modulation",
that>s just a different word. Just tell me "why would I want to run it".

[quote]sending the current seed of B is like sending B and all of A collected
upon it.
Well - if B is generated by a Pseudo-random generator, I agree. But the amount
of size reduction should be no larger than the size of the seed, I would believe.
why would that be? the collection count is based only on the carrier
entrpy.
[/quote]
No, the collection count is based on the number of "1"s in the B-stream.
That>s not the entropy.

[quote]Or, even different, why would I expect to find for a given source a good pseudo-random
generator and a good seed that would help me? This seems rather hard.

windows is shit again.
[/quote]
How does that relate to anything I>ve written?

So long,
Thomas

Industrial One · Guest

Ok, am I following up right: the user baseband information (the input
stream) when modulated makes it into a carrier stream, must have a 2:1
bias in order to allow compression later. But if the input stream is
random with a 50:50 entropy then modulating it would create 50% more
0s than 1s, so it would now have a 3:2 bias. Not enough? Do we re-
modulate which will now make the file twice as longer so it finally
becomes a stream with a 2:1 bias? How do you get the file to be
shorter than the original when it>s twice as larger now?

I knew what a state machine was the first time, I just thought you
meant something else. What is the role of this clock in relations to
the machine states? And what the hell is a FIFO?

Feel free to copy/paste parts of your previous posts if necessary. Am
still piss-ass drunk from last night (nothin' helps me sleep better
than alcohol ATM) so cant think straight.

This Valerian plant you suggested... list some stats, is it effective?
does it cause physical dependancy? it>s only available in europe and
i>m in north america, but I can get a couple friends from europe to
hook me up if possible, if all else fails, you wouldnt mind hookin me
up, eh?

jacko · Guest

Industrial One wrote:
[quote]Ok, am I following up right: the user baseband information (the input
stream) when modulated makes it into a carrier stream, must have a 2:1
bias in order to allow compression later. But if the input stream is
random with a 50:50 entropy then modulating it would create 50% more
0s than 1s, so it would now have a 3:2 bias. Not enough? Do we re-
modulate which will now make the file twice as longer so it finally
becomes a stream with a 2:1 bias? How do you get the file to be
shorter than the original when it>s twice as larger now?
the input stream makes it onto a carrier which has > 2:1 bias pre[/quote]
modulation. it is then moved on to its next state algorithmically to
be the next carrier state. which can itself be modulated ad infinitum.
why do you have the need to want to draw the carrier out into a stream
to be compressed later, when its compact representation is just so
much smaller. the carrier is always generated from its own past state
of less than 100 bits.
[quote]I knew what a state machine was the first time, I just thought you
meant something else. What is the role of this clock in relations to
the machine states? And what the hell is a FIFO?
clock to synchronize, first in first out stack.
Feel free to copy/paste parts of your previous posts if necessary. Am
still piss-ass drunk from last night (nothin' helps me sleep better
than alcohol ATM) so cant think straight.

This Valerian plant you suggested... list some stats, is it effective?
does it cause physical dependancy? it>s only available in europe and
i>m in north america, but I can get a couple friends from europe to
hook me up if possible, if all else fails, you wouldnt mind hookin me
up, eh?
valium made from it.[/quote]

Industrial One · Guest

On Jun 20, 7:39 am, jacko <jackokr...@gmail.com> wrote:
[quote]Industrial One wrote:
Ok, am I following up right: the user baseband information (the input
stream) when modulated makes it into a carrier stream, must have a 2:1
bias in order to allow compression later. But if the input stream is
random with a 50:50 entropy then modulating it would create 50% more
0s than 1s, so it would now have a 3:2 bias. Not enough? Do we re-
modulate which will now make the file twice as longer so it finally
becomes a stream with a 2:1 bias? How do you get the file to be
shorter than the original when it>s twice as larger now?

the input stream makes it onto a carrier which has > 2:1 bias pre
modulation. it is then moved on to its next state algorithmically to
[/quote]
How>s that possible when the input has a 50:50 entropy? There>s no
bias.

[quote]be the next carrier state. which can itself be modulated ad infinitum.
why do you have the need to want to draw the carrier out into a stream
to be compressed later, when its compact representation is just so
much smaller. the carrier is always generated from its own past state
of less than 100 bits.
[/quote]
So in order to compress our file to 77-100 bits, we just store the
machine state + how many ticks the clock would have to make to reach
the bit construction of our input stream. What is the process though?
What modifications to the input stream are made when the clock rises
or falls by 1 tick? Modulation just increases the filesize by adding
more 0s to it, but how do we bring the size *down* losslessly?

jacko · Guest

Industrial One wrote:
[quote]On Jun 20, 7:39 am, jacko <jackokr...@gmail.com> wrote:
Industrial One wrote:
Ok, am I following up right: the user baseband information (the input
stream) when modulated makes it into a carrier stream, must have a 2:1
bias in order to allow compression later. But if the input stream is
random with a 50:50 entropy then modulating it would create 50% more
0s than 1s, so it would now have a 3:2 bias. Not enough? Do we re-
modulate which will now make the file twice as longer so it finally
becomes a stream with a 2:1 bias? How do you get the file to be
shorter than the original when it>s twice as larger now?

the input stream makes it onto a carrier which has > 2:1 bias pre
modulation. it is then moved on to its next state algorithmically to

How>s that possible when the input has a 50:50 entropy? There>s no
bias.
[/quote]
the carrier needs the bias the input stream does not.

[quote]be the next carrier state. which can itself be modulated ad infinitum.
why do you have the need to want to draw the carrier out into a stream
to be compressed later, when its compact representation is just so
much smaller. the carrier is always generated from its own past state
of less than 100 bits.

So in order to compress our file to 77-100 bits, we just store the
machine state + how many ticks the clock would have to make to reach
the bit construction of our input stream. What is the process though?
What modifications to the input stream are made when the clock rises
or falls by 1 tick? Modulation just increases the filesize by adding
more 0s to it, but how do we bring the size *down* losslessly?
[/quote]
exactly on the store front. modulation will not increase the file size
as long as the carrier entropy is low enough. have you not seen that
the zeros in the input stream will be collected and not emitted when
the carrier is zero? the only modification to the input stream are
emit and collect and as long as the carrier entrpy is low enough there
is nett collection, hence shrinking input file, hence all input
collected.

cheers

Guest

On Jun 7, 8:05 am, The Jackal <j...@optusnet.com.au> wrote:

[quote]I am the Jackal and I will return with the story behind this
BREAKTHROUGH ANNOUNCEMENT in two weeks time!!!!
[/quote]
Well, it has been two weeks. What is your lame excuse this time?

jacko · Guest

earlcolby.pottin...@sympatico.ca wrote:
[quote]On Jun 7, 8:05 am, The Jackal <j...@optusnet.com.au> wrote:

I am the Jackal and I will return with the story behind this
BREAKTHROUGH ANNOUNCEMENT in two weeks time!!!!

Well, it has been two weeks. What is your lame excuse this time?
[/quote]
jackal ain>t me. If this was directed at me in the mistaken belief
then "Iza sorry massa. Ma bitch arse ain>t been fillin' your bank. Isa
bad slave massa, don>t hovel ma. Isa need ma walking cos ma rolls is
mashed"

Guest

On Jun 22, 2:08 pm, jacko <jackokr...@gmail.com> wrote:
[quote]earlcolby.pottin...@sympatico.ca wrote:
On Jun 7, 8:05 am, The Jackal <j...@optusnet.com.au> wrote:

I am the Jackal and I will return with the story behind this
BREAKTHROUGH ANNOUNCEMENT in two weeks time!!!!

Well, it has been two weeks. What is your lame excuse this time?

jackal ain>t me. If this was directed at me in the mistaken belief
then "Iza sorry massa. Ma bitch arse ain>t been fillin' your bank. Isa
bad slave massa, don>t hovel ma. Isa need ma walking cos ma rolls is
mashed"
[/quote]
And that read like a lame excuse to me still regardless who you may
be.

Instead of delivering clear instructions on compression, you try
writting some insulting language to get readers distracted.

And the question still remains: Cut out the garbage, WHERE IS THE
CODE?

Phil Carmody · Guest

earlcolby.pottinger@sympatico.ca writes:
[quote]On Jun 22, 2:08 pm, jacko <jackokr...@gmail.com> wrote:
earlcolby.pottin...@sympatico.ca wrote:
On Jun 7, 8:05 am, The Jackal <j...@optusnet.com.au> wrote:

I am the Jackal and I will return with the story behind this
BREAKTHROUGH ANNOUNCEMENT in two weeks time!!!!

Well, it has been two weeks. What is your lame excuse this time?

jackal ain>t me. If this was directed at me in the mistaken belief
then "Iza sorry massa. Ma bitch arse ain>t been fillin' your bank. Isa
bad slave massa, don>t hovel ma. Isa need ma walking cos ma rolls is
mashed"

And that read like a lame excuse to me still regardless who you may
be.

Instead of delivering clear instructions on compression, you try
writting some insulting language to get readers distracted.

And the question still remains: Cut out the garbage, WHERE IS THE
CODE?
[/quote]
Nonsense. The question is "why haven>t you killfiled the idiot yet?"

Phil
--
Dear aunt, let>s set so double the killer delete select all.
-- Microsoft voice recognition live demonstration


www.GetXFactor.com Leading Technology, Science, Agriculture News and information	Part of the Identityscape.com network... getxfactor.com jmoodmusic.com smartbusinesschoices.com mintdepot.com lowfaresalways.com evangelicalview.com shoppingpodder.com soproudlywehail.com webnews.ws currenthumor.com

Science and Technology news... Forum Index -> Compression Forum	Goto page Previous 1, 2, 3
Page 3 of 3