[FRIAM] Constraint Propagation and "Wave Function Collapse" Algorithm

Steve Smith sasmyth at swcp.com
Sat Oct 21 18:23:03 EDT 2023


SG> Just went down a 2-hour rabbit hole on the "wave function collapse" 
algorithm that emerged in graphics in 2016 but just
SG>  came onto my radar... Has anyone else explored it already?
SG> https://github.com/mxgmn/WaveFunctionCollapse

I think I 'get' from our myriad discussions about both dual-fields and 
bidirectional search why you got "rabbit holed" by this...

FWIW my "associative memory" self "squinting" from 100k ft saw some 
near-adjacents:

    https://en.wikipedia.org/wiki/Hashlife

         On one hand it is "memoization" up front, but it also has a
    possibility for a WFC style application for a *dynamic* landscape?

    https://en.wikipedia.org/wiki/Vector_quantization

         SOFMs seem like an apt near-adjacent to what you are maundering on?

    https://en.wikipedia.org/wiki/Attention_(machine_learning)

    The Text-Image transformers of DALL-E (et alia) seem to be a softer
    version of the constraint business?

I may be overgeneralizing or missed the focus of your fascination?

mumble,

  - Steve


> Many of you have written versions of constraint propagation algorithms 
> in one form or another. I like how this is framed by satisfying local 
> constraints with tiles (forward) and global constraints with overlaps 
> (backward propagation).  Of course, the name of the algorithm may be 
> metaphorical to QM as is its use of superposition for local stacks of 
> possible states, but I can't help wonder how Wheeler-Feynman Absorber 
> Theory or Cramer's Transactional Interpretation might be cast as 
> similar kinds of the same algorithm.
>
> more general applications:
> https://robertheaton.com/2018/12/17/wavefunction-collapse-algorithm/
>
> always like Dan Shiffman's Coding Train
> https://www.youtube.com/watch?v=rI_y2GAlQFM
>
> https://github.com/avihuxp/WaveFunctionCollapse#demo
>
> A nice interactive to get the feel for it:
> https://oskarstalberg.com/game/wave/wave.html
>
> A version in Julia :
> https://github.com/roberthoenig/WaveFunctionCollapse.jl/blob/master/usage.ipynb
>
> -S
>
> _____________________________________________________________________________________________________
> stephen.guerin at simtable.com
> stephenguerin at fas.harvard.edu
> Harvard Visualization Research and Teaching Lab 
> <https://hwpi.harvard.edu/eps-visualization-research-laboratory/home>
> mobile: (505)577-5828
>
> -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. .
> FRIAM Applied Complexity Group listserv
> Fridays 9a-12p Friday St. Johns Cafe   /   Thursdays 9a-12p Zoomhttps://bit.ly/virtualfriam
> to (un)subscribehttp://redfish.com/mailman/listinfo/friam_redfish.com
> FRIAM-COMIChttp://friam-comic.blogspot.com/
> archives:  5/2017 thru presenthttps://redfish.com/pipermail/friam_redfish.com/
>    1/2003 thru 6/2021http://friam.383.s1.nabble.com/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20231021/a9188d5d/attachment.html>


More information about the Friam mailing list