Hi,

 

Recently, I've been looking into the subject of information. After seeing a video on fractals I figured that a fractal is, in some sense, a dimension with missing pieces. The missing pieces can then be interpolated, if one likes, though without adding any new information.

 

The example I've been looking at is the function f:R->R, f(x)=sin(x).

It seems obvious that f being a periodic function would contain less information than some other function, say, sinc(x) (defined at zero to be equal to the limit).

It also seems logical that the function f(x)=0 would have even less information than both.

 

You could say that whenever there is less information spread on an entire dimension (like the x-axis), the limited information needs a rule/symmetry that would tell us how to fill in the gaps. In the sine's case we say that sin(x)=sin(x+2pi). Of course, one has to somehow count the information resulting from the law itself.

 

All this led me to the conclution that the more unpredictable a function is, the more information does it contain, so, noise actually has the most information one can get.

 

Maybe physical laws and symmetries are just a cover for the lack of information in the universe, or to an equivalent description - maybe the universe's dimensions are not full, but are actually fractals?

 

I'd like to hear any thoughts and comments you people have about this...

so, noise actually has the most information one can get.

Yup.

 

Of a simple thing like a string of characters, it contains more information if it takes more data to describe it. This is an okay introductions.

 

Maybe physical laws and symmetries are just a cover for the lack of information in the universe

Kinda. There's an idea in biology that a lack of information causes a default towards symmetry.

All this led me to the conclution that the more unpredictable a function is, the more information does it contain, so, noise actually has the most information one can get.

 

Correct. It takes more information to accurately describe noise. Whether the noise conveys anything or not is a different question. Incidentally, compression results in a string of data that looks almost like noise.

All this led me to the conclution that the more unpredictable a function is, the more information does it contain...

 

 

Yes. You would be interested in Shannon's information theory in which information is very much like entropy in statistical mechanics.

 

Information is like the "surprise" in a message.

How do you measure the "surprise" in a message/function? My first guess was that predictability has something to do with auto-correlation, but that doesn't seem enough. There is correlation and there is independence. Independence implies zero correlation, but zero correlation doesn't necessarily imply independence. I feel I need to measure the auto-dependence of a function to measure its predictability.

 

But how do I do that? How do I measure the auto-dependence (the "surprise") of a function?

I really forget the details now, been a long time since I looked at information theory. See if you can find some lecture notes online. Or find a good book.

Thanx, I will