Post your idiot spare time projects. i'm the subtle buttocks

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > Post your idiot spare time projects. i'm the subtle buttocks

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

PleasingFungus posted:

a* is super cool & fun because you can actually see it Do A Thing

it doesn't require advanced math or anything, it's just looking through the possibilities in an ordered way

here's a cool visualizer I ran into a while back that also has some variants; I really want to implement that guy's jump point search for something sometime

this is really awesome

# ¿ Nov 22, 2013 23:16

Adbot: ADBOT LOVES YOU

# ¿ Apr 24, 2024 09:59

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

coffeetable posted:

workingpeer i found your youtube channel

https://www.youtube.com/watch?v=dcrY59nGxBg

ps it's fantastic that you've made it so far with this sous vide thing

that video was amazing. the last 3 seconds alone make it worth watching.

# ¿ Dec 4, 2013 01:42

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

cheese-cube posted:

i know its ot but his latest vid is amazing https://www.youtube.com/watch?v=D5vVmwsLtCo

"there is no escape from the enemy who lives within" hail satan

"shot dead by nazis in a back alley in downtown vancouver"

rest in peace you sexy unibrowed gentleman

# ¿ Dec 4, 2013 18:32

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

in javascript, all variables are doubles or something. don't loving use a pow function to compute squares.

x*x computes in one clock cycle but pow(x,2) probably takes like 6.

maybe they check to see if the exponent has no fractional component, but i doubt it.

# ¿ Dec 15, 2013 21:53

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

i guess maybe if you're using javascript you're not too worried about counting flops

# ¿ Dec 15, 2013 21:54

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

fidel sarcastro posted:

counting all your projects would get pretty depressing.

i just live vicariously. i haven't had a project of my own in like a year :smith:

# ¿ Dec 15, 2013 22:08

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

Is it a Markov model or do you copy large chunks of text?

# ¿ Dec 28, 2013 21:45

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

Bloody posted:

A blend of second and first order markovs

next make a markov image generator using okcupid profiles pics to train the model.

# ¿ Dec 29, 2013 08:30

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

coffeetable posted:

O(n lg n) voronoi diagrams onna sphere

with this done i can adapt my climate model to use a) whatever mesh i want b) a multigrid solver

do you use great circles for your edges?

# ¿ Jan 3, 2014 20:47

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

i guess it probably doesn't matter because youre just going to integrate over the area anyway

# ¿ Jan 3, 2014 20:48

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

kwinkles posted:

verilog is gods own language

lol if your babby high level language cant specify what happens at each clock, thats barely even computering imo

going from cpu code to gpu code was awesome. you have to write so much closer to the hardware. i imagine going to an fpga is like that times 10.

# ¿ Jan 12, 2014 10:18

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

legos rule. stop motion is cool too

# ¿ Jan 28, 2014 00:29

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

suppose i wanted a single custom chip one time. we're talking just a couple grand maybe?

is that price following a moore's law pattern? will i soon be able to print out a custom chip for pocket dollars?

# ¿ Jan 29, 2014 01:45

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

coffeetable posted:

i quit my maths PhD to learn programming

same. now i make real money and learn more instead of making grad student money and barely learning

# ¿ Feb 6, 2014 10:11

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

coffeetable posted:

you end up somewhere that makes use of your maths knowledge?

yes, actually. I was doing image processing in the security industry

now im switching jobs to do computational geometry poo poo for a 3d printing/laser cutter company. contracting loving sucked, but the work was sometimes fun. hopefully new job will be even better

# ¿ Feb 7, 2014 00:58

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

old job is looking to replace me. if you are

a. american
b. dont mind las vegas
c. like fbi agents asking about drug history and foreign travel

then it was p good, i guess. pm me.

# ¿ Feb 7, 2014 01:18

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

homelan security and tsa only buy from american companies :911:

, so i only know other american companies

# ¿ Feb 7, 2014 01:37

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

pagancow your fan thing works good

# ¿ Feb 18, 2014 10:23

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

echinopsis posted:

us file avatar people just merge into one mega poster, with 13 shades of poo poo in every hour

it's hard enough with that renaming thread

shiiiit

# ¿ Feb 19, 2014 10:08

Adbot: ADBOT LOVES YOU

# ¿ Apr 24, 2024 09:59

Mathhole: Jun 2, 2011; rot in hell, wonderbread.

Sagebrush posted:

prepare for

basicaly there are two different fundamental ways of storing 3d data (not counting weird stuff like voxels): polygonal models and NURBS surface model. polygonal models are what you think, lots of flat surfaces that you move around by pushing and pulling vertices and with enough of them you can approximate a curved shape. polys are super lightweight and fast and they're easy to work with because you literally just grab a vertex and move it. video games and many movies use high res polys because of this (and because they don't care if your round machine gun barrel is actually a 12-sided prism)

NURBS modeling is a way of mathematically describing the curvature of a surface with a continuous polynomial equation. once you set the parameters for that equation, you can sample it at any point and get the exact value for the surface location, so a NURBS can be sampled at arbitrarily high resolution. it's the only way to 100% accurately record a curved surface like a sphere or the line of a car body in a computer and so is also how all modeling for engineering and design is done.

disadvantage of nurbs is that it's harder for people to wrap their heads around, because you aren't directly adjusting the surface shape but only controlling the equations that define the surface, so it can be harder to grasp initially. but it's the way to go if you want quality engineering models.

then within nurbs there are two main modeling paradigms: solid modeling and surface modeling. fundamentally they make the same type and quality of model but they're different mindsets. in solid modeling you act like you're starting with a block of wood and carving pieces away or gluing more pieces on until you get teh shape you want. in surface modeling you act like you're building a wire cage for the model and then stretching fabric or paper along the outside to get teh shape. ultimately surface modeling is the most open and flexible but also the most difficult to understand.

if u want a programming analogy, poly modeling is like python or something (used in industry but not for the really serious life-threatening stuff), solid modeling is like C++ (the bread and butter of 90% of the good engineering you see out there) and surface modeling is like assembly (total control over every aspect of your product at the expense of increased complexity)

rhino is primarily a surface modeler with the option to do solids and some polygon editing and conversion. solidworks is a solid modeler with a not-very-good surface modeler patched in and no helpful handling of polygon data. art apps like 3ds, maya, cinema 4d are going to be primarily polygon modelers with some basic nurbs surfacing tools thrown in and no solid-modeling paradigm.

so surface or solid modeling is where you wanna go for the highest quality and accuracy if you're making a model for manufacturing. HOWEVER there is one more aspect: the resolution of the 3d printer. most printers these days don't even approach the resolution of a quality cnc'd injection mold, so you can totally get away with a low res polygon model in most cases. in fact 3d printers take only poly models as inputs (stl file), but if you start with a nurbs you can render it out at low or high resolution as is appropriate for your printer and model scale. w/ polygons you're stuck at whatever number of polys you have.

essentially if what you're printing is warhammer figures or something and you don't care to go further then any 3d modeling program will do you. if you wanna start making useful mechanical parts then you want a real nurbs cad program most definitely.

this is a good explanation and i like you.

btw, my new job is writting computational geometry poo poo for a 3d printing company. its not spare time projects, but one day maybe ill write about a cool thing im doing

# ¿ Jul 12, 2014 04:25

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > Post your idiot spare time projects. i'm the subtle buttocks