Any details on AMD Leo demo?

Status
Not open for further replies.
Not practical for real time.

Well it seemed like it took a lot of graduates but after his dream discovery the filter seems to apply quite fast in that app in that ancient pc.

It's similar to what happens in biology, and other fields at first years of Ph'ds are needed but when radical discoveries stack it moves to months in undergraduate and then to a day in regular high school.
 
except fractal compression has been around for 20 years on the pc
it's had plenty of time to mature and become everyday the fact it hasnt tells you something
its lossy, using lossy compression to increase iq is ass backwards
 
except fractal compression has been around for 20 years on the pc
it's had plenty of time to mature and become everyday the fact it hasnt tells you something
its lossy, using lossy compression to increase iq is ass backwards
Did it seem lossy to you? it's infinite resolution. The fact that alternate approachs were taken does not mean its invalid.

A function that can define 2d patterns in fractal terms, can like hashlife, transform a function computation into a mere traversal problem, yielding the arbitrarily foward state of any arbitrary function.
 
Hard to tell from a youtube video, but the fact is it is a lossy compression scheme,
The program sort of says "A is the same as B so we can just store A and a little note as to where to duplicate B
and when it cant do that anymore to achieve the impressive compression ratio's often quoted,
it starts doing A looks similar to B so we can just store A and a little note as to where to duplicate B
and it fails hard when you have very little repetition in a texture or textures

as for "The fact that alternate approachs were taken does not mean its invalid"
is true but also true is "The fact that you think its cool does not mean it is valid"
 
Hard to tell from a youtube video, but the fact is it is a lossy compression scheme,
The program sort of says "A is the same as B so we can just store A and a little note as to where to duplicate B
and when it cant do that anymore to achieve the impressive compression ratio's often quoted,
it starts doing A looks similar to B so we can just store A and a little note as to where to duplicate B
and it fails hard when you have very little repetition in a texture or textures

as for "The fact that alternate approachs were taken does not mean its invalid"
is true but also true is "The fact that you think its cool does not mean it is valid"

Nope fractal duplication creates a perfectly nested bidirectional loop well defined and well behaved, each inner nesting performs a proportionally scale(continuous rate function ala derivative ala integral ala wavelet or fourier, but defined in terms of a few simple automata rules with self reference for specification and well definition behavior as well as ease of understanding without use of convoluted wording and procedures) indefinitely recursive traversal. Infinite frame interpolation and infinite detail interporlation via fractal wave decomposition(in laymans terms N-size kinematon automata or simply put a self-accelerated automata via self reference, traducing all to the basic universal computer that requires only to add and branch to be universal).

Or more simply a nested kinematon cellular automata interpolating N-size grid kinematon. I'm personally fond of the hexagonal implementation.


Think of height as being calculated in a continuous manner(in both time and space, intervals and extent or resolution, it has infinite spatio temporal resolution with continuous well defined behavior, and can replicate and extend any arbitrary pattern or function, in direction complexity and space parameters) as the cells fractalize into a higher N-cell state.

IT is mathematically equivalent to tessellation at Nth level

Infinite Zoom in and zoom out can be done whilst preserving perfect structure in both time and spatial domains


It is an application of gosper's algorithm to perform infinite speed arbitrary function traversal once it is properly specified in the kinematon model.


Building the perfect sphere does not increase in difficulty as the number of cells increases, the same procedure can be performed over and over faster and faster, and at N-state the kinematon properly defined state it becomes a perfect sphere that can be infinitely zoomed in or out of.


Remember NAND=NOR in functionality, all arbitrary complexity functions can be broken down into terms of a finite combination of nand statements. The kinematon need only two inputs
 
Last edited by a moderator:
Kinematon model or kinematics can perform very advanced procedures all N particle


It translated the 3d problem into the 2d domain and can scale to N particle N state in finite time


Arbitrary jump into any state of the simulation at N time in the future Past or in-between

Think of it as simulation perfect magnetic balls(without the messy fields, the continuous field has been translated into a finite 5 state automata self referent code)


The kinematon model applies to any domain and can scale to arbitrary complexity systems with arbitrary temporal and spatial precision(think of it as dynamic range, it has infinite dynamic range. It's an integer implementation of N-precision floating point arithmetic... it can handle quantum mechanics N-system N-precision in realtime in finite time via integer arithmetic)

PS

This solves a lot of problems, and can probably qualify for some records or something. In any case it does present problems and opportunities for copyright law legislation if you understand the extent of the application..
 
Last edited by a moderator:
Btw, there was preceding and groundbreaking work done of this field previously, had the great minds of the past not been lost progress would have been even faster.

"We have noticed in nature that the behavior of a fluid depends very little on the nature of the individual particles in that fluid. For example, the flow of sand is very similar to the flow of water or the flow of a pile of ball bearings. We have therefore taken advantage of this fact to invent a type of imaginary particle that is especially simple for us to simulate. This particle is a perfect ball bearing that can move at a single speed in one of six directions. The flow of these particles on a large enough scale is very similar to the flow of natural fluids."Feynman

Nothing made him angrier than making something simple sound complicated.

For Richard, figuring out these problems was a kind of a game. He always started by asking very basic questions like, "What is the simplest example?" or "How can you tell if the answer is right?"

The denial is never an option when faced with reality, even if the obstacle cannot be overcomed by logic it will be overcome when you transcend logic whilst remaining logically consistent.

The reasonable man adapts himself to the world; the unreasonable one persists in trying to adapt the world to himself. Therefore all progress depends on the unreasonable man.

George Bernard Shaw (1856 - 1950), Man and Superman (1903) "Maxims for Revolutionists"

In sum,

9 28 18 20 0 1
1 14 18 0 0 9

Well I'll go into some nice ultra simplifed examples of what's going on.

You can imagine a pattern as a raft a one d line being pushed by the addition. This is equivalent to a raft in a surface such as a cellular membrane, which moves along and interacts with other surfaces(all interaction and all information exchange occurs in 2d surface interfaces, and all 2d surfaces can be deconstructed into 1 d rafts)

Now this nice video showcases, what the raft model looks like in a biological cell

When you get to the raft portion, think of the nature of pattern as the structure of the raft and its potential interaction with the structure of other rafts including self-similar*(identical) rafts.

An even simpler example that illustrates the nature of the algorithm in laymans terms is an escalator, an escalator can move a static object of arbitrary precision arbitrarily up.


The object does not change, yet impressively this gives an extreme complexe pattern result as outcome.


PPPS

This also serves as arbitrary length realtime verification of arbitrary proofs.(very good to make sure your arguments make sense before making them by verifying them for logical consistency.)
 
Last edited by a moderator:
And of course we couldn't do without the moe representation of the algorithm for good measure.


edited darn youtube the vid could be viewed by some of the more prune, as NSFW but doesn't contain anything that should be so before any rational civilized person so I'm providing a link instead of embedding. Warned, if there is conflict with the link I will remove and replace with image.

Perfectly legal I assure.
 
Last edited by a moderator:
steamy, what in the hell are you talking about - seriously
fractal compression works look heres proof a guy pissing about with a pin toy
why dont you just get yourself a fractal compression tool or plugin and you will see it doesnt do all the things you think it does,
 
I have a serious question:

How much will it take to compress a 4MP photo with fractal compression without loosing the tiny details? How big will the resulting compressed image be? The current examples seem to be worked out the other way around by taking the algorithm, giving it paramters and ending up with something similar to what is expected but not exact duplicate of an existing thing.
 
@hoho, I dont think anyone has done a lossless fractal compresssion that would suggest it isnt that good
it depends how much similarity is in the photo
it works by finding a group of pixels and working out that if they are moved/scaled/rotated/transformed they match/nearly match another group of pixels, the storing that formula as you can imagine for a picture its billions of calculations and if there is very little repition in the picture the stored formulas became bigger than the data your not storing
 
@hoho, I dont think anyone has done a lossless fractal compresssion that would suggest it isnt that good
it depends how much similarity is in the photo
it works by finding a group of pixels and working out that if they are moved/scaled/rotated/transformed they match/nearly match another group of pixels, the storing that formula as you can imagine for a picture its billions of calculations and if there is very little repition in the picture the stored formulas became bigger than the data your not storing

FInite size formula very small merely repeated.

Again it's like saying a nand gate is some uber complex gate. IT is NOT.

IT is simple it is a PROVEN FACT THAT ALL FUNCTIONS BOIL DOWN TO ADDITION, and simple branching. YOU PUT THE NAND NEST IT WITH NICE BRANCHING CODE, and BY SPECIFYING THE ALREADY MENTIONED SIMPLE CELL KINEMATON EXAMPLE it behaves like a fluid or moving sand structure.

The branching determines the behavior of the fluid. AND it can jump with arbitrary precision to any point in the calculation. IT HAS INFINITE DYNAMIC RANGE BY DEFINITION. IT IS AN INTEGER IMPLEMENTATION OR SIMULATION RULE OF ARBITRARY FLOATING POINT VALUES,
 
1. finite size formula is not fractal compression, fractals are not fractal compression
2 no one is attempting to make something that looks like a fluid in motion, we are trying to compress images, why are you even mentioning random stuff like that
3.no one said a nand gate is complex, but when you have a cpu built from billions of them it becaomes complex
3 234x2700567859 do that calculation on a cpu that only has an add function then do it on a cpu with a multiply function which is faster ? simplicity is not the answer to everything
4. "The branching determines the behavior of the fluid" because the fluid id the result of the formula in compression you dont start off with a formula you start off with the end product and have to work out what formula will get you that final image (this is extremely hard) have a go take o picture of yourself and then come up with a formula that when iterated millions/billions of times produces that picture - you cant without the formula becoming massively bigger than the original data.

if you want us to believe your right post a photograph here uncompressed then post the same picture after it has been fractally compressed, we will compare then if you right we we all say "hail steamy for he is indeed the dogs bollocks"
untill then dont be suprised if people think you are sprouting random nonsense especiall when you start talking of crystals extended into the forth dimension
 
1. finite size formula is not fractal compression, fractals are not fractal compression
2 no one is attempting to make something that looks like a fluid in motion, we are trying to compress images, why are you even mentioning random stuff like that
3.no one said a nand gate is complex, but when you have a cpu built from billions of them it becaomes complex
3 234x2700567859 do that calculation on a cpu that only has an add function then do it on a cpu with a multiply function which is faster ? simplicity is not the answer to everything
4. "The branching determines the behavior of the fluid" because the fluid id the result of the formula in compression you dont start off with a formula you start off with the end product and have to work out what formula will get you that final image (this is extremely hard) have a go take o picture of yourself and then come up with a formula that when iterated millions/billions of times produces that picture - you cant without the formula becoming massively bigger than the original data.

if you want us to believe your right post a photograph here uncompressed then post the same picture after it has been fractally compressed, we will compare then if you right we we all say "hail steamy for he is indeed the dogs bollocks"
untill then dont be suprised if people think you are sprouting random nonsense especiall when you start talking of crystals extended into the forth dimension

Details details, look into it and you will see there is no disagreement.
 
I would think that expecially non photo-realistic rendering could benefit from very non- uniform sampling, e.g from say 4 - 1/9 samples per pixel if you have say detailed edges but fairly flat shading on surfaces.
The problem is determining whether a pixel will contain detailed edges ;)
 
Yup, it would require either a specific preprocessing step to do this desicion, which may very well in practice cost more than just sampling each pixel + MSAA, or that each frame is rendered in steps with increasing resolution, in a quad tree like way where there is a condition to decide to go to the next step or not.

edit: the disclaimer is that I have no experience at all in the field, so while its a decent idea in my head it will most likely not work at all.
 
Status
Not open for further replies.
Back
Top