/sci/ - Science & Math » Thread #11658058

275KiB, 1169x717, 01.jpg

View Same Google iqdb SauceNAO

Monte Carlo Geometry Processing

Anonymous Sun 10 May 23:58:04 2020 No.11658058 View Reply Original Report

Quoted By: >>11659021 >>11659312 >>11659959 >>11661453 >>11661572 >>11662640 >>11663002

>We reimagine geometric algorithms without mesh generation or linear solves. Basically "ray tracing for geometry"—and that analogy goes pretty deep
>Especially for problems with super complicated geometry, not having to mesh the domain provides a massive reduction in real world end-to-end cost. For instance, this model takes 14 hours to mesh and solve w/ FEM, but less than 1 minute to preview with Monte Carlo

https://www.cs.cmu.edu/~kmcrane/Projects/MonteCarloGeometryProcessing/index.html

https://twitter.com/keenanisalive/status/1258152669727899650

Holy shit!

Anonymous

Anonymous Mon 11 May 2020 00:09:51 No.11658096 Report

Quoted By: >>11658114 >>11660625

you have my attention

Anonymous

Anonymous Mon 11 May 2020 00:14:43 No.11658114 Report

Quoted By: >>11658732 >>11660193

>>11658096
if you want a summary read this thread:
https://twitter.com/keenanisalive/status/1258152669727899650

Anonymous

Anonymous Mon 11 May 2020 01:28:18 No.11658306 Report

Quoted By:

neat

Anonymous

Anonymous Mon 11 May 2020 04:16:37 No.11658732 Report

Quoted By:

>>11658114
duly appreciated for some nice content my anon

Anonymous

Anonymous Mon 11 May 2020 07:01:11 No.11659021 Report

Quoted By: >>11662870 >>11662944

>>11658058
Monte Carlo is almost always slower than any other method for doing anything. The only thing it is suited for is model fit verification.

Anonymous

Anonymous Mon 11 May 2020 09:33:28 No.11659312 Report

Quoted By:

>>11658058
cool, actual science.

Anonymous

Anonymous Mon 11 May 2020 10:19:08 No.11659385 Report

Quoted By: >>11659494

Someone eli5

Anonymous

Anonymous Mon 11 May 2020 11:40:04 No.11659494 Report

Quoted By: >>11660161

>>11659385
if you throw dots at insect you can work out what shape they are

Anonymous

Anonymous Mon 11 May 2020 15:32:33 No.11659918 Report

Quoted By:

Can't say I fully understand it but it seems like a new method to approximate certain linear PDEs with Dirichlet conditions on complex geometric domains? The only preprocessing involved is building the BVH tree structure to evaluate nearest-point calculations which is a lot faster than meshing a 3D domain. Anyone know how these BVH structures work? Can they be built quickly for moving geometry?

Anonymous

Anonymous Mon 11 May 2020 15:56:36 No.11659959 Report

Quoted By:

>>11658058
Interesting. So from my layman's understanding: instead of ray tracing in world space, the algorithm finds the boundary in surface space? Am I right or wrong? I would love to know more about this.

Anonymous

Anonymous Mon 11 May 2020 17:18:37 No.11660161 Report

Quoted By:

>>11659494
Isnt that just light

Anonymous

Anonymous Mon 11 May 2020 17:26:46 No.11660193 Report

Quoted By:

>>11658114
Well this doesn't make feel dumb, just ignorant of a lot of stuff.

Anonymous

Anonymous Mon 11 May 2020 19:53:28 No.11660625 Report

Quoted By:

>>11658096
And my depression.

Anonymous

Anonymous Mon 11 May 2020 20:49:11 No.11660850 Report

Quoted By: >>11660863

How is it raytracing?

Anonymous

Anonymous Mon 11 May 2020 20:52:05 No.11660863 Report

Quoted By: >>11661034

>>11660850
open the link and read the paper?

Anonymous

Anonymous Mon 11 May 2020 21:36:21 No.11661034 Report

Quoted By:

>>11660863
?????

Anonymous

Anonymous Tue 12 May 2020 00:03:31 No.11661453 Report

Quoted By:

>>11658058
bump cause interesting af

Anonymous

Anonymous Tue 12 May 2020 00:51:21 No.11661572 Report

Quoted By: >>11661600

>>11658058
>hurr muh CS doesn’t do anything important or interesting

Anonymous

Anonymous Tue 12 May 2020 00:58:47 No.11661600 Report

Quoted By: >>11661676

>>11661572
at least posts like yours make us seem mature and smart eh?

Anonymous

Anonymous Tue 12 May 2020 01:19:58 No.11661676 Report

Quoted By: >>11661813

>>11661600
it appears he was making fun of people who bash on CS for being codemonkeying.
I think you might be the slow one if you couldn't see that anon

Anonymous

Anonymous Tue 12 May 2020 02:11:44 No.11661813 Report

Quoted By: >>11661828

>>11661676
you're a retard who doesn't understand greentext

Anonymous

Anonymous Tue 12 May 2020 02:18:43 No.11661828 Report

Quoted By:

>>11661813
no i think unironically you're the retard here
anon's intent was clear

Anonymous

View Same Google iqdb SauceNAO figure12.jpg, 121KiB, 1438x370

Anonymous Tue 12 May 2020 06:32:57 No.11662632 Report

Quoted By:

Anonymous

View Same Google iqdb SauceNAO figure12.jpg, 121KiB, 1438x370

Anonymous Tue 12 May 2020 06:35:29 No.11662640 Report

Quoted By:

>>11658058
>For instance, this model takes 14 hours to mesh and solve w/ FEM, but less than 1 minute to preview with Monte Carlo
>visualization pi related

Anonymous

Anonymous Tue 12 May 2020 06:42:54 No.11662658 Report

Quoted By:

>using C++
i was expecting python at first but I guess it does make sense given they are computer sci

Anonymous

Anonymous Tue 12 May 2020 08:23:55 No.11662870 Report

Quoted By:

>>11659021
t. 19 year old data scientist

Anonymous

Anonymous Tue 12 May 2020 09:01:53 No.11662944 Report

Quoted By:

>>11659021
>Monte Carlo is almost always slower than any other method for doing anything
Then why is it used all the time in lattice field theory?

Anonymous

Anonymous Tue 12 May 2020 09:26:10 No.11663002 Report

Quoted By: >>11663043

>>11658058
This looks incredibly interesting, what fields / levels of math do I need to know to really understand what's going on in the paper, I'm assuming more than just differential equations?

Anonymous

Anonymous Tue 12 May 2020 09:43:30 No.11663043 Report

Quoted By: >>11663123

>>11663002
>differential equations
partial differential equation, vector calculus. There are a few big name equations in there that are also used in a lot of other fields of math and applied physics

Anonymous

Anonymous Tue 12 May 2020 10:16:30 No.11663123 Report

Quoted By:

>>11663043
Taking both of those next semester at Uni, thanks

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches