/sci/ why do engineers make a fuss about Fluid Dynamics? Is it

It's hard, but not the hardest.

DRAGONFORCE is the hardest.

File: 1459647430597.jpg (148KB, 1286x580px) Image search: [Google]

148KB, 1286x580px

>>7977096
Because you CS undergrads don't know how to do anything other than make pretty apps and websites.

Pic related is how you solve fluid dynamics problems.

>>7977128
"""""""solve"""""""

>>7977128
FEM a best

>>7977128
> pretty apps and websites
ok senpai, next time implement your own scientific simulation software :)

>>7977098
Could someone learn it by itself? I'm curious about it

>>7977146
Yeah, I do it on a regular basis.

CS fags can't code for shit.

File: sticker,375x360.u3.png (112KB, 375x360px) Image search: [Google]

112KB, 375x360px

>>7977149

von neumann would have solved it if he hadnt gotten brain cancer

he was quoted saying he thinks it is actually a lot simpler than it seems and he was showing interest in working on it

but his problem is he was very unfocused, since he was literally the smartest person on the planet in the last 100+ years, he couldnt just stay in one field, he had to innovate everything

oh well i guess we can just be grateful for what he did give us before he died young

>>7977096
in a nutshell, viscosity is a bitch
the euler flow is "easier" to deal with, but not always applicable, especially at high speeds.
It roots from the unsolved navier-strokes equations

>>7977096
fluid dynamics gets me hard >:)

File: Is9vjRJ.jpg (54KB, 430x760px) Image search: [Google]

54KB, 430x760px

>>7977149

File: typical cs fag on sci.png (287KB, 836x1065px) Image search: [Google]

287KB, 836x1065px

>>7977146
>clueless cs major thinks he's important

funny every time

File: hqdefault.jpg (12KB, 480x360px) Image search: [Google]

12KB, 480x360px

>>7977210

>>7977151
>He thinks it is actually a lot simpler than it seems and he was showing interest in working on it

Newsflash: Geniuses can be wrong, and there are no special insights accessible only to them that can't be approximated by fifty years of ordinary scientists and mathematicians making incremental struggling progress.

>>7977239
>>7977128
>mfw they only got algorithms
>mfw they've never tried anything challenging.
>mfw when computer science is the only field with problems interesting/worth solving
>mfw CNN's are so complicated no one knows how they work

>>7977096
>Fluid Dynamics
>engineers

So, lots of cum gargling then?

>>7977301
>>mfw CNN's are so complicated no one knows how they work
Combinatorics. Jesus christ you guys are retarded.

>>7977307
>combinatorics
>CNN's
What are you on?

>>7977314
Gradient descent is a local search technique you faggot. You're finding the *combination* of weights that results in the answer you want.

Fuck, just stop spreading your stupid.

>>7977316
Nice buzzwords. Still not combinatorics.
LeCun himself said he doesn't know why they work, you really think you do?

>>7977301
>they work

They don't, they suck.

>>7977319
What part of neural networks are a mystery to you? You're applying uncertainty where isn't any just to sound smarter than you actually are.

>>7977319
>>7977328
Also I mean, literally, what part don't you understand? As in this isn't a rhetorical question.

>>7977328
Pretty much the whole thing, can you explain why AlexNet is worse than ZFNet? Here's an easy one, what's the best way to decide on hyper parameters?

>>7977327
look up facebook's facial recognition research. Pretty much anything useful in machine vision is done with CNNs

>>7977337
That's implementation-specific. It's like asking why BMW's are worse than Audi's. The difference between the two has nothing to do with a fundamental understanding of *how* neural networks operate.

>what's the best way to decide on hyper parameters
That's another combinatoric design problem, much like the training process.

>>7977332
Literally the entire area is a mystery. It's a big black box, if you'd read anything more than a Wikipedia article about you'd know this.

>>7977359
I've implemented them and done things with them. I think I'm more than qualified to comment.

The whole
>hurr durr we cant no nuffin
doesn't really get us anywhere.

>>7977351
Yes, but you can explain why a BMW is faster than an Audi, or why an Audi has a fuel efficiency.

You can't explain why some CNNs work better, other than vaguely pointing out differences in the architecture that are apparent.

I.e the BMW is faster because the engine is designed to give more power. Pointing out that you use a pooling layer after every two conv layers, instead of one, doesn't explain why you get better results.

File: 1365292999904.gif (85KB, 540x786px) Image search: [Google]

85KB, 540x786px

>>7977359
>CS major livin' up to the memes

>>7977362
It's not a we can't know nothing. It's a we don't know why.

>>7977371
>>7977377
The two cars are different because one of the engineers at a particular company happened to stumble on a better design for the engine, while the other company didn't. Such is life with combinatoric optimization processes. Very rarely are they deterministic.

>>7977375
to be honest if you don't study computer science your just a waste of space at this point waiting to be automated.
See you at the singularity.

File: jesus christ.gif (1MB, 288x198px) Image search: [Google]

1MB, 288x198px

>>7977386
Oh god, lol. Please stop, you're killing me.

The fact of the matter is that everyone else is better at doing CS in their field than actual CS-majors. Face it, you don't know anything outside of time complexity notation and web development.

t. guy researching computational combinatoric design of mechanical parts

>>7977382
You're really hanging on to this "Combinatoric Optimization processes" thing. I'm not so sure if you actually know what that is. It doesn't apply here

>>7977393
Except that's what it is. If you dig a bit deeper, you'll realize that it's everywhere. Intelligence is combinatorial optimization. That's how neural networks learn. That's how evolution works. That's how people design things. You're exploring the possible solution space for good candidates.

>>7977392
>literal autism

your assesment of CNNs is getting cucked by your "experience"
if you "real engineers" would ever be able to transcend your tunnel vision you'd see that CS has more room for any innovation, scientific inquiry, or your beloved MUH PRAKTICUL APPLUCAYSHUN than putting an extra set of cylinders on an engine.

File: look at him and laugh.gif (2MB, 400x206px) Image search: [Google]

2MB, 400x206px

>>7977408
Lol, it's funny because it's the other way around.

I'm willing to bet I can program better than you. Can you apply your vast understanding of the laws of nature to bend it to your will? I'm going to guess not.

File: 2012-03-07_peter-norvig2.jpg (156KB, 1920x1093px) Image search: [Google]

156KB, 1920x1093px

>>7977386
I suggest that you examine this book very carefully before posting in this thread again.

>>7977412
>bend the laws of nature to your will

You realize the laws of nature don't apply to a fucking computer? Classic "real engineer" unable to think critically.

>>7977416
>Common Lisp
>not prolog

What is it 1970 again?
Also that's not the preferred book. You want the 3rd edition with Russel and Norvig

>>7977416
tits or gtfo

File: laughter.gif (2MB, 390x271px) Image search: [Google]

2MB, 390x271px

>>7977418
Yes, that's precisely my point. Your skills are subset of mine. Your field is trivial.

File: 1457852867418.png (2MB, 1186x1500px) Image search: [Google]

2MB, 1186x1500px

>>7977421
Yes, that's the one.

>>7977425
Gentlemen don't ask.

>>7977427
No you have a very broad set of skills that you're doing a very poor job of applying to a very specialized area.

All I read is this thread is an egineering guy talking about CS major topics (wchi he knows very little about).

File: image.jpg (67KB, 500x500px) Image search: [Google]

67KB, 500x500px

>>7977431
You a book from this decade... dare I say a modern one?

>>7977440
Yes a definite read for the pop science believers of impending singularity.

File: m'lady.jpg (11KB, 142x137px) Image search: [Google]

11KB, 142x137px

>>7977431

>>7977445
I can't tell if this is sarcasm or you're just stupid.

>>7977109
Power metal is great but it's definitely one of the less heavy subgenres of metal.

>>7977794
Technical death is better

File: 382f47e0.png (474KB, 648x654px) Image search: [Google]

474KB, 648x654px

>>7978272
I like them equally

they think solving PDEs makes their dicks bigger

it's the modeling that's difficult

even models are gross approximations with a shitload of assumptions.

god damn if you know this shit then it'll keep you up at night.

File: 1457558292640.gif (2MB, 500x750px) Image search: [Google]

2MB, 500x750px

>>7978563
working on this stuff made me lose a little bit of my mind tbqh

>>7977254
>Newsflash: Geniuses can be wrong, and there are no special insights accessible only to them that can't be approximated by fifty years of ordinary scientists and mathematicians making incremental struggling progress.

Only if P=NP

>>7978574
where were you studying?

I was looking at microfluidics before i switched to heat propogation. i always wondered what if i kept with fluids.

>>7977445
You realize that book has like 3 pages on the singularity?

>>7977096
engineers make CFD out to be hard as it is trying to find the right models and methods for the simulation you are doing, if you are doing subsonic flows you use different models to doing transonic, to supersonic, hypersonic, then there are dozens of random models that you can use for seemingly no reason, then there are dozens of settings for gas simulations and turbulence simulation, then if you go into the proper stuff, that isnt just, 'its sorta like this smudge' you need to have beefy as fuck computers. and thats just 2d cfd you go 3d and you many as well say goodbye to sense.

Essentially its just an engineers dream, an absolute clusterfuck that very few people understand, but they can charge money for.

>>7978604
Modelling for large eddy and Reynolds-averaged Navier-Stokes simulations using direct numerical simulations.

Pretty standard stuff.

>>7979329
It's like that because it is. There isn't one engineer that enjoys working with this patch work model mess with thousands of fudge factors thrown in.

>>7977109
2008 /mu/ called it wants its meme back

>>7979046
Defining what it is at the end of the book. Read the 25 chapters before that

>>7977128
Did you even read his question?