The result of an attempt to create Markov Chains.

-- 
https://ingoogni.nl

Post a reply to this message

Attachments:
Download 'markov_scene.gif' (309 KB) Download 'markov_scene.pov.txt' (3 KB) Download 'markov.inc.txt' (9 KB)

Preview of image 'markov_scene.gif'

ingo wrote on 28/06/2021 12:39:
> The result of an attempt to create Markov Chains.
> 

Interesting result.
I was used to see Markov Chains with nodes and arrows, but your image is 
visually prettier.

Paolo

Post a reply to this message

hi,

ingo <ing### [at] tagpovrayorg> wrote:
> The result of an attempt to create Markov Chains.

thanks, interesting.  like PG I've not seen Markov Chains used to generate
graphics/patterns, only knew of an application which generates "gibberish" text.


regards, jr.

Post a reply to this message

in news:web.60ddaa903c46af395e0fed26cde94f1@news.povray.org jr wrote:

> hi,
> 
> ingo <ing### [at] tagpovrayorg> wrote:
>> The result of an attempt to create Markov Chains.
> 
> thanks, interesting.  like PG I've not seen Markov Chains used to
> generate graphics/patterns, only knew of an application which
> generates "gibberish" text. 
> 
Oh, they are used for all kinds of things, page rankings, weather 
prediction (in a simple way). In generative music they are used for 
(beat)pattern generation and chord progression.

Here's an other one. Although I didn't feel like hand filling the chances 
in a 256 point transition matrix so I condensed it a bit.

I used a simple kind of domino like shape. Square with cutouts, 5 shapes, 
with rotations resulting in 16 shapes. 
Placed in a grid with random choise of shape but within rule set:
   - start bottom left, then progress row by row
   - left side of next shape is same as right side of left neighbour.
   - bottom side of a next shape is same as top side of neighbour below.
   - if the last piece is within the range of possible pieces, given the
     neighbours shapes increase its chance (in a naive way)

The Markov chains can also be used for texture generation. Instead of the 
simple shapes I used you can use square of textures with very simmilar 
forms.

An other application is adding vegetation to a landscape. Use some rules 
to select a range of vegetation types. Then use markov to select a tree 
from the bunch. A rule can be to have a look at the trees in the 
neighbourhood and then increase the chance on the same tree. The higher 
the order of the transition matrix the more "clumping" of the same type 
you get. 

-- 
https://ingoogni.nl

Post a reply to this message

Attachments:
Download 'domino2.pov.txt' (10 KB) Download 'domino2.png' (274 KB)

Preview of image 'domino2.png'

ingo wrote on 01/07/2021 20:00:
> in news:web.60ddaa903c46af395e0fed26cde94f1@news.povray.org jr wrote:
> 
>> hi,
>>
>> ingo <ing### [at] tagpovrayorg> wrote:
>>> The result of an attempt to create Markov Chains.
>>
>> thanks, interesting.  like PG I've not seen Markov Chains used to
>> generate graphics/patterns, only knew of an application which
>> generates "gibberish" text.
>>
> Oh, they are used for all kinds of things, page rankings, weather
> prediction (in a simple way). In generative music they are used for
> (beat)pattern generation and chord progression.
> 
> Here's an other one. Although I didn't feel like hand filling the chances
> in a 256 point transition matrix so I condensed it a bit.
> 
> I used a simple kind of domino like shape. Square with cutouts, 5 shapes,
> with rotations resulting in 16 shapes.
> Placed in a grid with random choise of shape but within rule set:
>     - start bottom left, then progress row by row
>     - left side of next shape is same as right side of left neighbour.
>     - bottom side of a next shape is same as top side of neighbour below.
>     - if the last piece is within the range of possible pieces, given the
>       neighbours shapes increase its chance (in a naive way)
> 
> The Markov chains can also be used for texture generation. Instead of the
> simple shapes I used you can use square of textures with very simmilar
> forms.
> 
> An other application is adding vegetation to a landscape. Use some rules
> to select a range of vegetation types. Then use markov to select a tree
> from the bunch. A rule can be to have a look at the trees in the
> neighbourhood and then increase the chance on the same tree. The higher
> the order of the transition matrix the more "clumping" of the same type
> you get.
> 

A starting point for a lot of new ideas... thank you,

Paolo

Post a reply to this message