/sci/ - Science & Math

>>12303630
In general, there have been cool developments in geometry and geometric processing:
https://www.cis.upenn.edu/~jean/DIFGEO-ppm.pdf

>>12303630
begone back to /g/, cs-monkey

>>12303670
>back to /g/
but math and CS aren't /g/
they'd just say to go back to /sci/

Manifolds are becoming more common topics of study in CS
https://www.visgraf.impa.br/cma2011/

>>12303681
/g/ is the CS board, unironically

>>12303692
>unironically
Not in my experience. They care about basic programming questions and some interview stuff but they don't really give a shit about CS.
You can't talk about graphics or compilers there, much less complexity theory.

Barely anybody on /g/ actually knows anything about CS

>>12303697
yea, you're right
/g/ is full of code-monkeys whose biggest achievement is installing some obscure linux distro and bragging about it there

"Many cognitive processes involve transformations of distributed representations in neural populations, creating a need for population-level models. Recurrent neural network models fulfill this need, but there are many open questions about how their connectivity gives rise to dynamics that solve a task. Here, we present a method for finding the connectivity of networks for which the dynamics are specified to solve a task in an interpretable way. ..."
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1008128#abstract1

>>12303712
lol yeah. no hate to codemonkeys, they usually seem happy with the work they have. It's just that I wish more people really knew what CS was about, especially on this board

Do people really not think CS doesn’t have math?

>>12303630
https://arxiv.org/abs/1209.5993
Mathfags will ignore this because "muh CS"

>>12303908
where do start with Geometric Complexity though. It's not even in the Arora Barak book. And is it related to Computational Geometry or are the names just similar.

I've been working as a programmer for about 5 years since i did my BCS, are there any cool new developments in computer science?

>>12303937
>where do start with Geometric Complexity though
Be comfortable with a good amount of Arora-Barak, be comfortable with abstract algebra up to at least galois fields. Literally most match you learn will be useful for geometric complexity theory, but you don't have to study every single thing out there initially. There are two good resources you can use to start:
Ikenmeyer's course and lecture notes:
http://pcwww.liv.ac.uk/~iken/teaching_sb/summer17/introtogct/index.html
Landsberg's book:
https://www.cambridge.org/core/books/geometry-and-complexity-theory/15E3ABA3FF14E1054574663F60250D80
> It's not even in the Arora Barak book.
It's because that book is a survey of classic results from the 70s-90s with the PCP theorem added in.
>And is it related to Computational Geometry or are the names just similar.
The names are similar but it's closer to algebraic complexity theory and algebraic geometry.

>>12303946
Yeah, lots of interesting stuff regarding graphics and some stuff in learning architecture. What types of things interest you?

>>12303953
thanks anon, I'll check it out. although I guess I'll need to brush up on my "non-cs math" first

>>12303955
I did a specialization in graphics and animation, I did a special project on raytracing and I'd love to know abotu any specific developments tehre.

but anythign outstanding really. I'll read any articles on provability of algorithms, optimization, any new algortihms in general, any new unsolved problems. I get a heavy feed of industry trends like ai, machine learning, raytracing, big data, NLP, whatever, but not much more abstract knowledge in the CS realm. what am i missing?

>>12303970
For graphics, the CMU research group always has really really cool results:
http://graphics.cs.cmu.edu/
For algorithms, I thought this was cool:
https://arxiv.org/abs/1906.10340
>>12303966
haha I get what you mean, but I slowly learned that there's very little math out there that couldn't be part of CS. It just ends up being a question of motivation. Everything from differential algebra to functional analysis has its place in CS - hell, algorithms researchers have been using quite a lot of functional analysis in their work, especially when working with low distortion embeddings from high dimensional spaces to low dimensional metric space representation.

>>12303630
A theory or B theory?

>>12304072
Doesn't matter, both are cool. I recently read through some of Gallier's logic for proof systems book, so I've been on a Theory B kick

>>12303712
do you think /g/ is a big reason why /sci/ doesn't like CS?

most people don't know about actual cs because the barrier to entry between it and what people usually think cs is (programming) is much much higher.
it's basically 'pure' maths.

>>12304218
There's a decent amount of theoretical CS that isn't too hard to learn out the bad. I think the *most* interesting and most impactful parts of CS are definitely hard and close to pure mathematics, but I don't think it's too hard to appreciate basic compiler theory as it is, say, geometric complexity like in >>12303908

>>12303717
Super interesting. Thanks for posting. I'm skeptical about the use of RNNs and NNs in general for theory building though, especially when you compare work done with them to, say, Bayesian models or RL. Do you have any other examples where you think the use of neural nets actually led to some novel insight about the brain or psychology?

>>12304262
>I'm skeptical about the use of RNNs and NNs in general for theory building though, especially when you compare work done with them to, say, Bayesian models or RL
Yeah, I am too. Ultimately if you tread like a careful Baysesian and don't take RNN's and NN's too seriously in theory past their ability to do efficient search, it works out. Proper ML theory is really cool though.
>Do you have any other examples where you think the use of neural nets actually led to some novel insight about the brain or psychology?
Look up "The Eighty Five Percent Rule for optimal learning" from Nature. It should be open access

honestly I like listening to podcasts and the like about simple physics subjects, but even as a CS undergrad I find the little I can grasp of computational physics so boring. I like compilers better, too bad I'm probably never going to work on a compiler anyone will ever use. I have interest in AI too, but I'm even more of a layman when it comes to it

>>12304291
>I find the little I can grasp of computational physics so boring.
You need to be motivated to understand the physics for computational physics to be exciting. It's a cool field and one I'm sort of interested in, but I think the parts of CS that end up contributing to physics theory (hamiltonian complexity comes to mind) are most interesting to me
>I like compilers better, too bad I'm probably never going to work on a compiler anyone will ever use
you never know, compiler engineering and compiler theory are pretty well motivated and active fields
>I have interest in AI too, but I'm even more of a layman when it comes to it
You can pick up a few textbooks and do experiments with libraries. It's a wildly competitive field, but it's not exactly inaccessible - you would have to find a professor to work with you though.

>>12303670
>>12303681
>>12303692
>>12303697
>>12303712


Was /g/ always meant to be a consoomer circlejerk? You'd think there'd be more acceptable overlap with /sci/ topics

>>12304611
Dunno, but that’s what /g/ is. What’s sad is that the CS vs. EE thread has more traction on /sci/ because it’s easier to know nothing about CS and just hate it than it is to understand it and see what’s interesting about it.

/g/ cares mostly about basic programming shit though. They tend to scoff at anything that takes more mental power than USE HASHMAP, which is partially why I think the CS major hate is rampant here.

>>12304218
this

https://cstheory.stackexchange.com/questions/2011/complexity-of-linearized-ising-model-at-0

https://arxiv.org/abs/cond-mat/0309240

CS + physics is really cool

Redpill: if we can respect physics as a science CS is math and science.

What do you think about the recent advance on TSP approximation?

>>12306418
Could you link a paper or article about it?

>>12306418
interesting if not particularly useful.

Gallier's work is relevant to this thread
https://www.cis.upenn.edu/~jean/home.html

His upcoming book "Differential Geometry and Lie Groups A Computational Perspective" looks interesting, it's on libgen in the draft version.

>>12306663
Gallier's work is really cool. I've been meaning to go through his parametric pseudomanifolds paper

>>12303720
You see a lot of pseudointellectuals assuming computer science is software engineering.

>>12306668
yeah, you're right. unfortunately, it's most of this board...
not to say CS undergrad doesn't have its problems, but I don't know how someone could seriously look at CS research and think it isn't full of serious math and serious engineering.

>>12306684
It definitely is, I've just finished my CS undergrad and am more than halfway through a mechatronic engineering undergrad.
We mainly covered algorithm theory, cryptography, discrete math and I elected to specialise in artifical intelligence topics like search algorithms and machine learning.
I don't even know how to make a program from start to finish, computer science is a branch of discrete math.
I'm hoping to do my masters in it sometime after I finish this engineering degree.
I have to write a thesis on mechatronics next year and I'm stumped for ideas, I want to try and make it as computer science related as possible as I honestly find that much more interesting. I'm thinking maybe something that can incorpoerate graph theory or machine learning into a mechatronics topic, like optimising search algorithms for autonomous vehicles with A* algorithms (which I realise has already been done, but something in that line of thinking) - like this.
https://youtu.be/qXZt-B7iUyw

>>12306705
>computer science is a branch of discrete math.
I would argue a lot of CS is starting to incorporate continuous and other nondiscrete math into its core theory these days.
>incorporate CS into mechatronics
I'm sure you've read motion planning literature right?

>>12306705
Also damn, great video

>>12306663
>>12306666
First edition is up too now

>>12306745
>motion planning
We covered it in my comp-sci course but surprisingly not so much in mechatronics.
So I know the theory but applying it is a little fuzzy, I have until February next year before I need to start working on the thesis so I have some time to get more familiar with it.
I remember learning A*, D* and probabilistic roadmap.
Most motion stuff we've learned for mechatronics are more concerning robotic arms for assembly lines, which is still interesting - but the course droles on in more depth than I care about the fine details of screws, nuts, material properties and types of DC motors.
My brain is definitely more wired for the pure sciences.

>>12306705
>>12306830

Based power combo desu, but what motivated you to also study mechatronics? Employability?

Perhaps for your mechatronics thesis you could look at using algorithms and ML to improve and optimize the pathways of stuff like robotic arms and self-driving vehicles in a way that saves energy/costs or strains the machine less.

Do you do much CAD/M in your mechatronics degree? And if so, what, if any compsci disciplines helped with that?

>>12307022
Thanks, I've jumped around degrees a bit. I'm doing a double degree, but I've completely finished my CS major.
I originally went into mechatronics right after highschool because I've always been into robotics. Then I picked up cognitive science because I realised I'm pretty interested in the brain/ pure logic/ epistemology. Then I realised it made more sense for me to study computer science which covers a fair bit of that but actually has job prospects, and I enjoy programming.

But yeah basically I've always thought that automation will be the next industrial revolution so to speak, and that it would be ideal to try and cash in on it - and I won't lie, my 17 year old self partly just thought "mechatronics" sounded cool.

Using ML learning for optimisation was one of my first ideas, I had the idea to make ‘smart’ power outlets to manage your usage with machine learning, only to find out that that’s already a product you can buy. Another idea I'm toying with, inspired by a math assignment I just finished where I had to create a system of ODEs representing the current in the human heart: maybe I could use machine learning to detect anomalies in heartbeats, I think that might be how some smart watches do it. But I also don't know that it would be necessary, there might be more effective traditional control systems for that.
I saw an interesting paper recently for a CS masters thesis where this guy attempted to recreate the falcon 9 landing system with machine learning in a python simulation.

As for CAD, I only had to do one unit on it. Which is a bit of a shame because I actually enjoyed it
I'd say it probably helped a little in knowing a lot more coding for MATLAB than they expected us to. But we only did relatively basic CAD, designing basic machinery and running some simple simulations with them. I still have a year to go though so we'll see.

>>12303641
Information geometry (which is where stochastic geometry is likely heading towards)

>>12307188
what's cool about gallier's work is that he shows that continuous / classical mathematical structures aren't irrelevant when talking about computer science. Same with Keenan Crane's work - both of these guys end up studying differential geometry in a way that uses both continuous and discrete ideas to study geometry and geometry processing at large.

Here's a fun little thing to go through for ray tracing and graphics. The math isn't hard per say (thought it can quickly get hard when you look into the literature at large), but ray tracing is fucking cool.

https://raytracing.github.io

>>12304145
Not really /g/ in particular, more like Silicon Valley and game development getting conflated with an idea of what computers are meant to be and the general conception that universities train people to get a job associated with what they study, which all together melts into some general idea of CS being something that has to do with (commercial) software development instead of computation in general.

>>12303692
/g/ is the consumer electronics board.

>>12308577
That's true, and that CS programs are mostly compliant with this image because they get $$$ from both the students tuition and funding from the companies that want more codemonkeys in the market.
It's bad even at the masters level. Only when you start doing a PhD do you get exposed to real CS at most places.

Question. Does it take a solid graduate-level background in randomized algorithms, cellular automaton or computational geometry to really master procedural generation? I'm enamored by the possibilities of procgen and am currently working it into a video game, but I only have a bachelor's degree and feel like the approaches I am taking are especially naive.

Pic related: it's my current archipelago generator.

>>12308888
>solid graduate-level background in randomized algorithms, cellular automaton or computational geometry to really master procedural generation?
no, but studying hard material wouldn't hurt

>>12303630
http://www.cs.cmu.edu/afs/andrew/scs/cs/15-463/2001/pub/www/notes/fourier/fourier.pdf

>>12303670
>>12303681
They're right, CS is /sci/

>>12308888
Btw there is some proven in Unciv https://github.com/yairm210/Unciv

Bump

>>12303760
most CS grads don't pay much attention to the more theoretical courses and once they leave uni all that theoretical shit is irrelevant to code monkeying.

>>12310118
This is the only fair variant of the "cs isn't math" argument I've heard. But its an ad hominem since the issue is "most cs grads" not cs itself.
I think its the same reason science majors think engineers are stupid, its because a lot of them are. But that doesn't mean engineering is a stupid pursuit - I'd say that the best engineers may be on par with the great scientists, many of the big names were both.

Its not entirely math, sure, but it largely is a subset of math.

>>12310118
what if I don't give a shit about most "cs grads" aka software engineering majors? What if I care about CS?
>>12310270
>Its not entirely math, sure, but it largely is a subset of math.
This is true, but the parts of it that are math are incredibly close in culture and content to pure math. I don't think anything from >>12303908
would strike someone as being "not math."

>>12304221
>>12304218
What is 'pure' mathematics?

>>12303630
What can I do to get a CS research internship if I have no prior research experience? However I have read CS research papers before and implemented one of them. It was a neural network.

>mathematical
>cs

Help me brainstorming: Is there anything where I can combine plants/gardening with algorithms/statistics?

>>12311143
in the institution i wont disclose we do computer vision with aerial photography for crop prediction, and computer modelling coupled with thousands of measurement stations for active pest control, that allows farmers to cut down the use of pesticides 8-10 fold. and i'm aware of maybe 5% of the stuff that we actually do here.

>>12311192
Cool.
I was actually asking about something I can do on my balcony with my mom.

>>12307948
why is P+N?

>>12311192
monsanto/bayer?

>>12311432
nah government founded, frying people with 5g, vaccines causing autism, microchips controlling your brain. i already lost the track of what is classified and what's not, so i'm keeping a tight lid

>>12311412
It's in the first book here, chapter 8 on diffuse materials, section 8.1: https://raytracing.github.io
"There are two unit radius spheres tangent to the hit point p of a surface. These two spheres have a center of (P+n) and (P−n), where n is the normal of the surface. The sphere with a center at (P−n) is considered inside the surface, whereas the sphere with center (P+n) is considered outside the surface. Select the tangent unit radius sphere that is on the same side of the surface as the ray origin. Pick a random point S inside this unit radius sphere and send a ray from the hit point P to the random point S (this is the vector (S−P)):"

Here point p identifies vector P between a point within the material to its surface (I think? I haven't read through all of these notes), and n is the normal with respect to p. Then you do diffusion calculations based on (P + n)

>>12310670
There's no hard line, but it ends up being largely cultural. Mathematics that largely exists in motivation to understand other mathematics, though that ignores all the pure math that came from other disciplines (Noether working in physics and getting pure algebra results comes to mind). Lots of theoretical CS goes both ways here - results in quantum information has solved things in mathematical physics and operator theory and algebra (Connes embedding problem was given a negative answer by MIP* = RE). Geometric complexity theory is largely problems in algebraic geometry, representation theory, and parts of topology.

>>12303630
This is cool as shit:
https://arxiv.org/abs/1401.3916

>>12311432
CIA has for decades used satellite imagery to determine crop yields across the world. This is important in agricultural negotiations and gives the US a major edge.

>>12306668
lots of pseuds study CS in undergrad hoping it's software engineering.
all of the other pseuds studying other stuff see that and conclude CS is nothing but dumbfucks.

They do this while quietly important all the research that has gone into computer vision, learning, robotics and motion planning, numerical analysis, bioninformatics, computational physics, and the theory work in communication and say "lol CS has done nothing for academia"

>>12310847
>CS isn’t mathematical
Lmao fuck off undergrad

>>12310676
Ask a professor if you can work with them. A masters is needed at least for industry research, PhD preferred for better labs and other positions in both academia and industry. What fields are you interested in anon?
You can also work for labs in biophysics and computational physics as a software engineer (proper engineering and research) but you should beef up your resume with experience. What’s your background?

What do you guys think of pic related? I am skimming through the third edition. Reading the first chapter so far there are two things I found very interesting. One was when he explains why he chose to use an assembler language instead of a higher level language for the book. Something like: it may seem trivial to save 1 millisecond, but if you had to run a calculation a trillion times, that 1 millisecond saves 11.6 days of computing. I found that kind of mind blowing to think about.

The other is that he says later in the book, there is discussion of an algorithm that can determine whether white can always win a game of chess (assuming no mistakes are made) but that it would take such an unfathomably large amount of time to run that we will certainly never have an answer in our lifetimes.

I wonder, would that algorithm be basically just the brute force method of considering every possible game of chess? Or something more interesting?

>>12312633
You didn't post the picture my dude

>>12303630
>share interesting things that are /sci/
HoTT - since Voevodsky ripped, it isn't visible nearly as much as it deserves to be.
Lawvere theories - very lovely, also doesn't get nearly enough attention, especially among code monkeys.
Obviously, post-quantum complexity theory has been pretty big in CS, such a shame that perhaps the most visible guy here is Aaronson, the eternal brainlet.
>>12308888
>procedural generation
Dead end for the purposes of your pic. You need to do simulation to get anything even remotely realistic. Your best bet if you want "infinite" maps relatively quickly is meme learning.

>>12313199
>HoTT
I remember HoTT being memed on here in the same way that category theory is (great subjects, but not for those without mathematical maturity and experience), but you're right that I don't see as much love here or outside of this board for type theories in general.
>Lawvere
I remember reading about some of his work in regards to the relationship between UTMs, the halting problem, and Godel incompleteness:

https://cstheory.stackexchange.com/questions/10635/halting-problem-uncomputable-sets-common-mathematical-proof/10636#10636

>quantum complexity
quantum Hamiltonian complexity and quantum information research in the CS departments has been really great. I'm really interested in the former and I personally want to learn more about entanglement in the latter
>Aaronson
I dislike the dude because he's the face of "my personality trait is nerd" and because he always seems to be the one with the most to say among a group of people who know way more than him. There are better computer scientists (Watrous, Gharibian, Aharonov come to mind) in the same field as him who work with both the CS and physics but aren't nearly as haughty. I don't know - something about him really puts me off, and it makes me cringe thinking that CS, particularly the part of it dealing with quantum information, is most associated with that dude.

>>12312633
What book?

>>12312546
I'm from a third world country. Study at a good-to- medium level university and am strong academically. Nothing very special about me

>>12313785
bump

>>12313785
>>12314232
Isn’t the answer obvious then

>>12314253
(. Could I at least get an interview or something? I have had minor research experience in my own university.

This might be a brainlet question but I'm trying to understand elliptic curve cryptography.

How does one find suitable parameters for elliptic curves in order to make the cryptography secure? For example in Bitcoin's secp256k1 they chose "p = 2^256 - p^32 - 977". How does one find such parameters? Did they just brute force a lot of prime numbers and compute the order of generators in the EC groups or is there a deeper reasoning behind it?

>>12313374
>I remember HoTT being memed on here
I vehemently memed it on irc. Don't remember much about it being mentioned here. It's the perfect topic for autists though, univalent foundations are absolutely yummy.
Type theory in general is something nobody on /sci/ knows anything about, so it's understandable you can't find many posts about it here.
TT autists are too busy playing around to shitpost and TT isn't really an undergrad topic so that further reduces its popularity here. You can notice that only undergrad stuff is being talked here, with the exception of occasional yukari autist and his CFT wank shitpost.
>quantum Hamiltonian complexity and quantum information research in the CS departments has been really great
Yes absolutely, finally CS being useful outside of code monkeying. I'd make it mandatory for grads to learn the basics, it's an essential field in modern CS.
>Aaronson
I get the same feels about him. He's pretty much perfect image of the stereotype about Dunning-Kruger in CS. Can't stand that guy, especially when he opens his loud mouth and gives his brainlet takes on string theory.

>>12314267
Professors are usually looking for PhD candidates. A strong interest and also perseverance are the qualities they are looking for. After all doing a PhD is a long trek. Also see
https://sciencecareergeneral.neocities.org/

>>12303689
>Manifolds are becoming more common topics of study in CS
How does CS benefit from understanding manifolds? Genuinely curious.

>>12313199
>Lawvere theories - very lovely, also doesn't get nearly enough attention, especially among code monkeys.
Please tell more.

>>12314757
>How does CS benefit from understanding manifolds? Genuinely curious.
Graphics processing, geometry, and dimensionality reduction. The UMAP algorithm in particular uses a fair amount of topology:
https://github.com/lmcinnes/umap/blob/master/README.rst
There are more uses, but these are the big ones.

>>12314383
>Yes absolutely, finally CS being useful outside of code monkeying.
I mean, there was already CS that was broadly useful for without being about code, but I agree with your sentiment - CS, after 50-60 years of its academic childhood, has become a big boy subject
>I'd make it mandatory for grads to learn the basics, it's an essential field in modern CS.
Ehh I don't know if it's strictly essential, since many people, even in TCS, probably wouldn't work in it. I like quantum Hamiltonian complexity a lot, but it's niche, and it takes more than a decent familiarity with both physics and CS to jump into.
That being said, I think the experience with mean field theory alone is enough to justify CS grads to learn it - mean field theory is just largely useful, regardless of context.
>gives his brainlet takes on string theory.
he also called geometric complexity theory the string theory of CS - something that's hard to get into and hasn't given complete results of what it wants to explain. Personally, I think he sort of grifts from one popular meme topic to another. Like, among physicists and computer scientists who always look to solve hard and interesting problems, Scott always seems to pick out the most reddit problems. I know that seems cheap, but it's the best way I can describe it. "waaaaaah what would happen to quantum computers if we had closed timelike curves!!!"

That being said, he made the following amazing shitpost image in pic related

>>12303630
This is great:
http://www.cs.cmu.edu/~15850/handouts/matousek-vondrak-prob-ln.pdf

>>12308577
what's funny is that ME and EE majors for the longest time bragged that they did less science and got more money, and now they're salty because they see brainlet CS majors in mediocre programs doing less than they are and getting even more money.
CS is great, not only because the subject is intense, but because even at its lowest levels, it tramples on the economic self esteem of engineering students, who by in large are concerned with their ego

>>12307948
I really dislike this book. The way the code is written in this book is ugly, unwieldy, and in many cases it will simply run forever, SPECIALLY the "generate random vector in a sphere" part, which he does by generating random vectors and discarding those who dont fit. And if you just want theory, the book gives very little, always putting it back. Absolutely dont recommend it

>>12315398
>"generate random vector in a sphere" part, which he does by generating random vectors and discarding those who dont fit
Holy shit. This can't actually be real.
Please tell me he knows how to sample n-sphere in non-brainlet way. This is undergrad lingebra.

>>12303630
bump

>>12303692
partly

>>12316852
not really, it's the software dev and consumer electronics board

>>12314280
> Did they just brute force a lot of prime numbers and compute the order of generators in the EC groups or is there a deeper reasoning behind it?
it's a combination of both. You use the theory to rule out bad choices, then do some crunching / simulations to get numbers that seem to work well, and finally write the AES standard or something similar. Friendly reminder that a lot of number theory research starts off as simulacra, sphere packing experiments, etc etc..

>>12312330
no one thinks CS has done nothing for academia you fucking faggot. people see CS as a largely disorganized, foundationless, heavily financially motivated field with some pieces of legitimate academic research and others that are absolute memes like ML and quantum computing.
>>12311833
CS goes both ways here - results in quantum information has solved things in mathematical physics and operator theory and algebra (Connes embedding problem was given a negative answer by MIP* = RE).
This is a meme. No serious theoretical physicist gives a shit about quantum information theory. Non-physical theory, uninteresting to particle physicists and quantum field theorists. Mathematical physics is barely physics.

>>12317338
>people see CS as a largely disorganized
The divide between theory and systems sure, but what else?
>foundationless
Wat. The foundations are in intuitionistic logic, algebra, and combinatorics.
>heavily financially motivated field
Not any more than your average field of engineering or science. Come back when we don’t have physics researchers in the pockets of medical device research
>No serious theoretical physicist gives a shit about quantum information theory
Lmao what? I guess Zurek and Dieks aren’t serious theoretical physicists? QIT has been great for research in decoherence and entanglement
>non physical theory
This really comes down to your view between classical information and thermodynamics. Physicists usually come together and say the following: it doesn’t really matter because the results it produces work when we treat information in a physical way
>uninteresting to particle physicists
This is blatantly untrue if you look at and of the quantum information conferences and the HEP-th people discussing their work. Yes, the non-holography people too.
>mathematical physics is barely physics
LMAO
Okay, say goodbye to a large chunk of modern fluids research, every nice part of quantum mechanics due to functional analysis, and the main source of PDEs in physics research today.

It’s so painfully obvious you’re an undergrad in physics.

>>12317338
>No serious theoretical physicist gives a shit about quantum information theory
Quantum information theory was established and is studied mostly by serious theoretical physicists.
Also
>no one gives a shit about
Condensed matter physicists have like ~25 years of results using quantum information to talk about gapped Hamiltonians. You're actually full of shit.

>>12318216
>>12317593
Please, don't bully the LARPing feeble-minded highschoolers. Ableist pigs.

>>12303630
Does machine learning count as mathematically CS? This Fourier Neural Machine solving parametric PDEs are record speeds got me interested.

https://arxiv.org/abs/2010.08895


All things considered can this make cum render faster on Blender?

https://playground.tensorflow.org/

pretty neat

>>12313050
>>12313480
Sorry lads, completely forgot. Pic related. Also, it was microsecond not millisecond.

>>12303692
As much as /sci/ is about math and not racial science and IQ.

>>12318272
Machine learning theory certainly does. There are some really cool results from that field, but they are drowned out massively by experimental research and development of engineering frameworks for use on practical problems.
It’s fine to want to solve practical problems, but machine learning is in dire need of more theorists now that the technology exists

>>12318476
They’re good reading but not necessarily for learning from. They’re an amazing encyclopedia of our algorithmic knowledge between the 70s and late 90s. It’s a great source of knowledge.

That being said, there have been major advancements to both algorithms and other parts of theoretical computer science in the last 20 years. In particular, there’s way more abstract algebra in mainstream CS, with analysis following up (analytic combinatorics comes to mind) and topology getting a lot of focus due to its relevance in both theory and high dimensional data analysis.

>>12318598
Thanks, I was concerned about it being outdated on some of the topics, seeing how old it is and how fast CS develops, but I wasn’t sure. Which books would you recommend for a novice? My schools CS program was mostly software engineering so I feel like I’m missing a lot.

>>12318722
I assume you already know calculus
>prelims
first chapters of Knuth's concrete math
any intro to proofs book
linear algebra by Hoffman and kunze
>algorithms
CLRS
MIT OCW
https://jeffe.cs.illinois.edu/teaching/algorithms/book/Algorithms-JeffE.pdf
>Automata
Sipser for a gentle introduction
Soare's Turing Computability if you want to learn logic proper
>compilers
programming language pragmatics (for the basics)
engineering a compiler (to get into nitty gritty)
Sorry anon, I actually don't know much about compilers past what intersects with type theory
>complexity theory
Arora-Barak
The Complexity Theory Companion


These are the basics that every undergrad should know about theory. There's way more stuff after this, but you'll need to stock up on more pure math.
In particular, if you like algorithms, I recommend you do lots of probability theory - eventually measure and analysis becomes important there, but you can get away without doing that at first given many of the immediate combinatorial probability applications. This is a great text:
http://wwwusers.di.uniroma1.it/~ale/Papers/master.pdf

>pure math
>probability

jk lol I hate that meme.

>>12320133
>>12318979

>>12320133
>measure theory isn’t pure mathematics
Kill yourself

test

>>12320133
>you’re gonna need pure math
Lol there’s way more than probability that you’re gonna need for TCS.

I have three positive variables x, y, and z, with the constraints that x+y+z = 1. If I want to feed x, y, and z into some generic optimizer, how do I transform them so they can take on any value between 0 and 1?
Do I just normalize it, so I do x = x_in / (x_in+y_in+z_in)?

>>12317107
So, yes really.

>>12303630
support the math and give a care
TWO THUMBS UP

>>12320275
Depends on what function you're trying to optimize.

For the people in this thread who are doing some kind of advanced math, how was your father/teachers? I majored in business at Uni because I thought I was stupid and minored in computer science (22 credits) when I realised I wasn't actually getting an education. I want to try and apply for a technical master. I'm thinking about data science or software engineering. What would advise me guys?
More importantly how do I get rid of this crippling fear that I am too stupid to tackle complicated topics?

>>12320275
>any value between 0 and 1
Literally impossible.
Computers can only handle a countably infinite number of values.
The "real" number line has uncountably infinitely many values, though.
Even between 0 and 1.

>>12321204
What's the generic approach?
>>12321755
fuck off

>>12318272
This is pretty cool

>>12318593
You're right, but the theory is gigabrain stuff.
A good amount of the people making ML papers and software don't fully understand it.

>>12317593
>intuitionistic logic
disgusting.
Platonism is true, constructivism is cope.

>>12322027
>the generic approach
There isn't one. It literally depends on what function you're optimizing.
Are you working with a differentiable manifold? Then a generic approach would be the method of Lagrange multipliers.
This isn't neccessarily the most generic approach even on differentiable manifolds, but it's very simple and fairly computationally friendly.

>>12303670
>>12303670
/g/ is just fucking tech junkies. CS isn't just hurr-durr I LOVE TOYS

>>12322597
>t. brainlet who relies entirely on contradiction proof

>>12314757
>How does CS benefit from understanding manifolds? Genuinely curious.

Reimannian (and even non-reimannian) optimization requires a deep knowledge of topology. For example, trees are better embedded in the Poincare hyperbolic model than in the regular Euclidean space, so, embed heirarchical data in the hyperbolic space by optimizing over some loss.

>>12321348
>More importantly how do I get rid of this crippling fear that I am too stupid to tackle complicated topics?
Do homework with people, study for 2 hours a day, review material often.

>>12303937
Computational Geometry: Algorithms and Applications

>>12323935
Geometric complexity theory != computational geometry

>>12322846
Any good resources?

Is discrete math useful for anything other than coding? Specifically processor design? I’m a computer engineering major with a lot of free space in my schedule junior and senior year and was wondering if that would be useful.

What's a good way to get into TCS research or really any mathematically heavy stuff that is related to CS if I am at a third world shithole? All they do is produce papers about muh-sheen learning, and even those papers are trivial applications of it, or just plain uninteresting (e.g. ooga-booga native dialect with NLP applications).

>>12325639
Unironically combinatorial optimization techniques are used in VLSI design.
http://www.dm.uni-bonn.de/~held/publications/bonntools_montreal.pdf

>>12326069
See the parts of theory that you can use to study NLP, and then branch out.
There’s plenty of interesting machine learning theory out there. That, and NLP can always benefit from research in efficient searching, which hits both applied and theoretical notes when you start trying to search for increasingly difficult structures

>>12322838
>t. brainlet who can't see the forms

>>12326715
>t. brainlet who hasn't advanced past theory of forms

>>12314757
>How does CS benefit from understanding manifolds?

I'm not sure about CS in general, but the manifold hypothesis (the idea that real world high dimensional data exists on a lower dimensional manifold) is a big part of ML theory.

>>12326293
That’s super interesting. Are there any other math or CS topics that would be useful to me with VLSI and FPGA?

>>12303760
From what I've seen "Computer Science" in the States is poorly named; should be called Software Engineering.

In my country we have Software Engineering (which is widely popular) and "Computer Science" which is much more theoretical and less commonly known.

I was browsing /x/ and came across pic related, I'm pretty sure it's completely fabricated make-believe since I can't find anything with relevant search terms (including the references at the bottom), but the idea of a program like this still interests me. How would I go about making a toy automata that tries to mimic what Logic Soup does? My guess was to derive properties from the commutator of A and B (there's slightly over 43 million possible commutators for 4 by 4 boolean matrices), but I have no idea what these properties should be. I'd assume I'd also have to track other information like position, field influence/radius, etc. Thoughts?

>>12328296
It depends on the program.
Not shitting on software engineering - it has its own can of worms - but you're right in that there's a mismatch of priorities between the academic and the industrial here.

>>12328438
It might vaguely be something like Conway's Game of Life. See https://youtu.be/xP5-iIeKXE8 and https://youtu.be/C2vgICfQawE

>Math
>CS

Lmao. CS Profs literally have to dumb down their materials for undergrad CSlets, so they'll be able to pass (see the Cinderella Book by Ullman).

>>12329918
I don't really care about CS undergrads. I care about the material prior to its dumbing down, aka mathematical CS.

>>12328296
The culprit really is English there.
German "Informatik", French "Informatique", Italian "Informatica" or Japanese "情報科学" do a much better job at describing the field they encompass.
Software Engineering is but one of many specializations you can choose later in your studies.
There is a massive mismatch between the English-speaking world and the rest of the world with CS.
The US abusing the academic system to create "code monkeys" also plays into that.

>>12331657
Information is very important in CS, but there's a way to talk about computation without informatic content. Computer Science in the US is horribly misnamed at the undergrad level (it should really just be called codemonkeying), but Information Science also fails to capture what CS is about all the way through.

>>12329918
Who gives a fuck about CS undergrads

>>12332063
That again comes down to America not having an in-company education system organized by the state.
You don't fucking need to attend a uni or college to become a code monkey.

The Japanese name indeed only is "information science", but the other languages incorporate mathematics into the name.
In English it would be Informathic.

>>12307058
>recreate the falcon 9 landing system with machine learning in a python simulation

'Interesting' would be an understatement to me. Can you link the paper? Would it technically be considered a 'Computational physics' simulation? CFD?

Thank you for the insight.

>>12307058
>I saw an interesting paper recently for a CS masters thesis where this guy attempted to recreate the falcon 9 landing system with machine learning in a python simulation.
Link to the paper please

>>12333403
I think there are two paths to CS becoming better and more standard: either it commits to being like physics and being closer to pure mathematics in undergrad, or it goes for the engineering route where they blitz you with a wide array of knowledge from the big subfields, but leave you to investigate a subfield by yourself.
I think the latter is more tenable in the US given the interests of CS undergrads even though I'm more invested in the latter. That being said, CS undergrads would bitch about the program getting harder as is.

>>12336375
>either it commits to [...] being closer to pure mathematics in undergrad
The situation is probably different in the US, but here in continental Europe, it's pretty much that.
It's way easier to graduate in CS without programming even once than to graduate without having proved some serious maths on your own.

>>12321348
>More importantly how do I get rid of this crippling fear that I am too stupid to tackle complicated topics?
imo unless you are a literal mental retard, no one is too stupid for anything. how smart you are is purely about how much work you put in. abilities that are well developed in smart people grow with practice. all you need is time. all the famous megageniuses in science worked their butts off nonstop to get there.

Is computer science a science or math? computer science is supposedly not about computers, so what is actually being studied scientifically and what conclusions are being drawn from empirical data? isn't it all just math?

>>12336517
>how smart you are is purely about how much work you put in
low iq post
>>12336532
It's computer science. They don't study nature so it's not science. They don't elaborate mathematical structures or theory, otherwise it would be math. They study computation, algorithms and computer architecture. All of these explicitly deal with computers. The "not computer" part stems from TCS being divorced from coding and software development, as well as computer engineering. But, everything they do even in the abstract relates back to making machines do computations. All of the complexity theory shit, the quantum information theory pseudery, the learning theory, all of that nonsense is fundamentally geared towards, founded in, and designed to be implemented with computers. It isn't science for this reason. Physics doesn't have engineering as an aim nor as a basis for it's theories. Fluid dynamics and quantum electrodynamics are not grounded in the constraints of extant electronics or machinery they are as broadly applicable to nature as possible. Mathematics literally does not concern itself with anything but purely abstract mathematical structures, it's utterly irrelevant that those structures be computable, physically relevant, or "logical" in the way that most people would term them.
>what is being studied scientifically
nothing. they don't hold themselves to the standards of physical science, if they did all of their work would reduce to materials science, quantum electrodynamics, particle physics, and statistical physics, because that's what computers actually do. the most serious research done on computation is done by statistical physicists.
>what conclusions are drawn from the empirical data
they don't work with empirical data anymore than mathematicians do.
>is it all jsut math
they do very little if any real math considering they don't actually develop pure mathematical theory. there are no structures in math that come from computer science

>>12336509
>It's way easier to graduate in CS without programming even once than to graduate without having proved some serious maths on your own.
This is my experience as well.

>>12336549
Most of what you said was incorrect for a number of reasons, but your biases / ignorance is clear. However
>there are no structures in math that come from computer science
Wait this is completely false though. Expanders, relativization, tropical semirings, reductions, cubical foams, etc. are all pure mathematics from CS.
This isn’t even to mention the work in algebraic geometric from algebraic complexity, in functional analysis from low distortion metric embeddings, in finite fields from cryptography, and the absolute fuckton of combinatorial research that CS has put out there.
>the most serious research done on computation is done by statistical physicists.
Lmao, reversible computation is something every researcher knows about period. The most serious intersecting physics and computation research is being done by condensed matter physicists and computer scientists on classifying Hamiltonians. The Ising model is important but only one example from which people talk about quantum Hamiltonians or mean fields.

You’re legitimately full of shit, kill yourself

>>12336549
>if computer science were rigorous it’d focus on particle physics in computer systems
You must be baiting, nobody can legitimately be this retarded.
Learning about junctions and doping is important for computers, but it’s not really what makes them fundamentally interesting, even to physicists. Physicists have traditionally contributed to computer science in ways that affect core computational theory. The “shit” like algorithms and complexity have attracted researchers from all disciplines for a reason.

You talk about computers like you’re an engineering major and it shows lmao. Ironically engineers end up knowing the least interesting and most rote parts related to computers in the first place.

>>12336532
The foundations of the field is in mathematical logic and combinatorics. The late 90s were spent introducing (abstract) algebra into the core theory, which was a MAJOR success, and the 2000s and 2010s have been spent either augmenting to current research or expanding horizons, especially to topology, algebraic geometry, and functional analysis.

It’s a young field, but it’s considered a core mathematical science for a reason. The only thing worth laughing at it is the ridiculous amount of shit undergrads because CS education in the US is a joke until the PhD level.

>>12336549
n-nooo you can't do maths in CS, that's illegal!!!1!

>>12336640
The guy you responded to will move the goalposts. He’ll say something like “no computer scientist has solved a big open problem in math.” And if you tell him that a CS PhD and his advisor solved Kadison-Singer, a problem in functional analysis that came from quantum mechanics, he will seethe even more and move the goalposts further back.

You cannot win with these people. They have decided CS is trivial and not worth studying, regardless of whether that’s true or not.

>>12336659
no need to go even that far. The recent MIP*=RE would also make him seethe

>>12336602
>>12336618
you're misinterpreting, any abstract theory that does not reduce to physics is not useful or meaningful to a good physicist. Computation is interesting in so far as it pertains to physical limits to information processing and the possible exotic behavior that the hardware can exhibit. That's it. There are no other components that are informative to physical theory. If CS was a science that would be the whole of what they studied. The rest is a kind of auxiliary pseudo-mathematical framework geared towards one end, making better computation machines, better software. The rest is cope. I don't like mathematicians, and I don't appreciate their schizophrenia but they really do not care about the applied end of things and this is admirable. they have the luxury of doing so by construction. CS does not. It is tethered to actual physical machinery in the real world. This means that everything done that purports itself to be science will be done by physicists and by CS people trained in physics. Nothing else would suffice in order to earn the title of science.

Also, condensed matter physics is really not the state of the art of current physical theory. It is probably the most mundane and least interesting of all the productive subfields of the science, and i'm not sure why you faggots wave the flag so enthusiastically over such a boring topic.

I thought you niggers were good at logic. If by construction physics studies the physical world nad it's laws, then every discipline claiming to be a science must necessarily bend to the laws and procedures of physics. There is no escape from this.

>>12336724
Nobody was saying CS was a science. They were disputing two claims
>CS hasn’t contributed any mathematical structure
There are numerous examples as to how this is false.
>the serious parts of CS would have to be the particle physics inherent to the construction of the computer
We can sit here and talk about our opinions on quantum information theory, or really statistical mechanics vs. classical information theory, and we would get nowhere. Instead, on a practical note, I will add that computer science has lent surprising credence to physical theory as of late if you look at the hep-th papers on holography and beyond. That, and the theorems from quantum information are both immediately useful in understanding entanglement and in the upcoming field of implementing devices that use various methods (trapped ions, anyons and tqft, etc etc.). The physicist doesn’t really care if information is physical - they care if information can yield results about the physical. It can and it has.
>Also, condensed matter physics is really not the state of the art of current physical theory. It is probably the most mundane and least interesting of all the productive subfields of the science,
This is the dumbest fucking take I’ve ever heard. You can be in high energy and appreciate condensed matter. Honestly, if fractional quantum hall doesn’t interest you, I seriously doubt your place in physics and think you’re just chasing meme science than actual physics.
Threadly reminder that condensed matter theory doesn’t need to have pbs documentaries and black science man in order to rally public funding ;^)

>>12336724
>Computation is interesting in so far as it pertains to physical limits to information processing and the possible exotic behavior that the hardware can exhibit. That's it.
Ah yes, yet another physics undergrad who thinks they have a grasp on <stem field> despite having a popsci tier understanding of it.
If we’re talking about just the limits here, there’s still a massive body of work needed to be done on Ising spin glasses and NP-hard problems, not to mention fleshing out QMA problems. On the side of purer interests, problems in complexity touch a lot of other fields in mathematics. P vs. NP is on the cutting edge of algebraic geometry right now, and MIP* = RE had implications for operator theory and functional analysis.
>he rest is a kind of auxiliary pseudo-mathematical framework geared towards one end, making better computation machines, better software.
Ah yes, the old “I flipped through sipser and thought this is what the field is about”
>CS does not. It is tethered to actual physical machinery in the real world. This means that everything done that purports itself to be science will be done by physicists and by CS people trained in physics.
CS is not about real world machinery - it’s always been agnostic to the realizability of its results, even from the very start with the digital computer. There’s a big body of work dealing with computability in more complicated structures, such as the reals or arbitrary rings. These are analyzed with models and ideas that are not immediately realizable. Computer science, as a field of mathematics and investigation, does not tie itself down to machinery. The important thing is that these ideas almost always end up being realizable despite the concerns of physicists (get fucked boomer physicists, there have been papers addressing noise models for decades in quantum computing).

Don’t you have an undergrad electromag test to be studying for? Shoo, the adults are talking

>>12334325
>>12335042
Here you go.

A Robust Control Approach for Rocket Landing - Reuben Ferrante

https://project-archive.inf.ed.ac.uk/msc/20172139/msc_proj.pdf

I would call it computational physics yeah, though he's simplified it down to 2D in Box2D.
I've got it bookmarked for inspiration when I start writing my own, still not set on a topic yet.

Picrelated is a question on my first-year undergrad Data Science test. Was my answer to the question or my choice of University the mistake?

>>12337385
Both.

>>12337385
Your answer was more or less correct given the choices. The Turing test, otherwise known as the imitation game, which like the Turing Machine, is more or less Turing’s definition for something that hadn’t been made rigorous yet. In this case, that definition was basically “what does it mean for a machine to be intelligent?”
Searle destroyed his argument with the Chinese Room though

>>12337445
I wasn't really happy with the answer I selected, but of the options I was given, it seemed closest to the truth. The given "correct" answer is a meaningless statement IMO.

>>12337445
I'll read up on the Chinese Room. Hadn't heard of it.

>>12337459
Your answer is meaningless.
Wheter machines differ [from each other? from humans? from kiwi?] in their capacity to imitate [do humans imitate intelligence? or what is the reference here?] intelligence [undefined buzzword].

>>12337472
As for your first question, I'd argue its implied that the difference is between other machines. I accept your point though.

>>12337445
>Chinese Room
"Mindless symbol-pushing" seems like an apt metaphor for illustrating the problem with that flawed approach to mathematical study which prompted Lockhart's lament, and which leads to farcical situations like >>12335960.

>>12337445
Yes but most researchers do not really care about the Chinese Room experiment. That is, they don't care whether a machine is truly intelligent or not as long as it delivers results. The Chinese room thing is more of a philosophical discussion

>>12337492
>>12337493
Yeah of course, this isn’t to knock on machine learning or artificial intelligence. These avenues are both very interesting and important - the Chinese Room just highlights the difference between our idea of intelligence (ie, the semantic vs syntactic value) and the ability to produce “accurate” answers. Some of my interests are in ML theory - we almost all care more about probability theory and math being able to solve hard questions on robust domains, not creating an “artificial brain.”

>>12337445
I genuinely don’t understand the Chinese Room. How do humans “understand” something in a different way than a system.

>>12337385
Macquarie???
I did this unit last year.

>>12337814
>Macquarie
Other side of the planet matey. Sorry to hear this has happened at least twice in the world.

>>12337794
The point of a Chinese Room is that algorithmic manipulation of symbols is not the same as deriving semantic meaning from those symbols.
There is a distinction between making an algorithm that can push symbols with human meaning and observing that a machine can actually assign semantic value to those symbols. No matter how effortlessly the person in the Chinese Room can give correct answers, or at least seeming like he knows Chinese, simply looking up phrasebooks or even having a script of answers is not evidence or proof of understanding or thought.

The algorithmicist, or generally the computer scientist's role, exists to study problems in a semantic setting, ie using mathematics. Their role in writing down the algorithm is figuring out how to 'solve the problem' without having to reference any semantic value.
Stuff like this is why computers are invaluable proof assistants but need a metatheory of judgements in order to be able to help prove anything at all. Don't take this as a dig at machine learning or AI research - read >>12337638 as to why. It's just in our usual sense of computing, machines do not and cannot think just because they can perform tasks with apparent accuracy.

>>12336640
this kills the "only maths and physics researchers do real maths" fags, aka most of /sci/. I understand defending maths as being largely...well, a mathematician's activity, but the way this place worships physics and gives exceptions for physics into the cool kid mathematics table is very weird.

Just saw this ramble from a math grad student on cryptography, what do you guys think?
>i just think the cryptography literature is fake deep autism that gets off on the application of elementary pieces of abstract algebra and alg num theory bc those things are somehow more intrinsically interesting than the more sophisticated analytic and combinatorial machinery used in real TCS ofc u can learn about whatever crap about abelian varieties to do supersingular elliptic curve diffie-neurodivserse-hellman but it's mostly pseudointellectual fake deep autism that provides nothing new of substance to the mathematics to existing algebro-mathematical discursive whereas the contributions of niggas doing probabilistic or combinatorial animals that draw on algebraic / integrable animals feed the animals endogenously and bidirectionallyrather than from the low iq blackbox application of some codified yellow book "look mommy look what i learned in school" autism

>>12338817
Cryptography is applications of algebra, algebraic number theory, and theoretical computer science to security and other associated things. It's weird because it sort of only contributes to its own theory most of the time, but at the same time cryptographic application is involved enough that it justifies its own study.
It sounds like your friend has spent a lot of time hitting the books for his quals and is constantly asking himself to value mathematics by its ability to comment on other mathematics. That's an important thing to have on the face of your career, because you should always contextualize your work, but you end up missing the big answer: it's fun to study and is motivated in some way or another.

This friend sounds like he'd have similar things to say about operator theory and functional analysis in physics. Something along the lines of "it's given results, but it's fake deep autism that by itself doesn't offer much to mathematics until mathematicians make sense of it." It's a sentiment that doesn't really mean much other than "I'm salty that research doesn't always go in the direction I'm personally motivated for."

>>12336704
Nobody tell him about analytic combinatorics lmao

>>12303670
Back to /pol/, race baiter

>>12337908
Damn yeah, they must recycle questions.
Interface looks identical too.

>>12336992
Thanks anon

>>12336724
>condensed matter physics is really not the state of the art of current physical theory.
This is bait. Both cmp-th and cmp-ex have regularly exciting results that are of interest to both physics and engineering.

question for the thread: which one of impagliazzo's worlds do you believe we live in?

>>12313199
>HoTT - since Voevodsky ripped, it isn't visible nearly as much as it deserves to be.
hott was originally motivated by practical applications and then everyone who wanted to work on those applications ended up working on hott itself instead.
this obviously makes it seem like a useless circlejerk to the outside world, even to the applied TT world.

>>12313374
>I don't see as much love here or outside of this board for type theories in general
/dpt/ has a bunch of type theory nerds that lurk but practically never post about it, unless someone else does. you can probably bait them out if you get lucky.

>>12340201
"Pessiland" seems like the most likely outcome.
It's the least convenient one, so you can practically bet on it.

>>12327831
There is no advancing past the truth.
There is only delusion.

>>12340201
>>12340655
i believe in cryptomania, mostly because i cannot imagine that someone will suddenly come up with a magical way to break hashing or public key cryptography for regular turing machines.

>>12340201
>/dpt/
Yeah you’re probably right, but I find that the programming threads on /g/ are a huge mixed bag
>which world
I have no immediate reason not to believe in cryptomania. I hope pessiland isn’t true though lmao

>>12340897
>but I find that the programming threads on /g/ are a huge mixed bag
they are, and honestly /dpt/ isn't worth lurking in. it's very rare that you find something interesting there.
the majority of it is programmers with <5 years experience discussing programming languages.

>>12315953
uh oh. I got an A+ in my undergrad linear algebra course and it's not immediately obvious to me how to sample points on an n-sphere in a non-brainlet way. pls elaborate?

PS should i look into switching schools?

>>12341178
>how to sample n-sphere
It's not too hard to understand, just look it up and understand the best ways to do it.

>>12341178
1) sample n+1 variables from N(0,1)
2) divide the resulting vector by its L2 norm
ok i lied, you need to understand normal distribution and its properties too, then you can put together how this particular transformation of Gaussian is uniform over n-sphere
but still, an undergrad should figure this out, it's more intuitive than sampling cube and rejecting points that don't fit and it scales better to higher dimensions

>>12341534
>2) divide the resulting vector by its L2 norm
All you did was deformation retract the a point in the n-box onto the n-sphere ;^)

>>12340965
I know /g/ is mostly shitposting, but I feel like there's a major amount of anti-intellectualism there. Most people scoff at you if you don't have a desire to maximize profit and minimize academic effort. Obviously there's a balance, but I don't really want to business logic even if it pays well, and I shouldn't have to feel the wrath of tech nerds in that field for not wanting to do so.

>>12304611
it's been that way since like 2012. rip /prog/

>>12342169
rip
fuck /g/, they even have a brainlet CS thread up there too: >>>/g/78696787

>>12342169
New programming thread up on sci

>>12303630
DID SOMEONE SAY MATH IN cs
https://www.irif.fr/~jep/PDF/MPRI/MPRI.pdf

>>12337385
>Exam administered through bullshit online multiple choice test

What kind of trash tier CS program is this?

>>12343215
also a horrible question with a horrible answer lmao
the Turing test is not interesting to computer scientists other than "let's make something that could pass it"

>>12340201
>this obviously makes it seem like a useless circlejerk to the outside world, even to the applied TT world
HoTT is so cool though, and it was literally motivated by the practical problem of mathematics putting its money where its mouth is. Voevodsky got scared that math was approaching unverifiability seeing as how his paper and many others were just accepted into annals without being properly examined, and in fact they had errors.
Last I heard, they had found a way to reasonably realize most of HoTT computably

>>12343881
it's still not obvious that hott will actually help with interactive theorem proving, compared to basic dependent type theory.
i'll believe it when a hott theorem prover has a large body of formalized mathematics.
for now, to me it seems like you could also build automation to transfer stuff along isomorphisms after doing some initial work in properly hiding away implementation details, and it isn't obvious to me that the hott approach won't just be similarly laborious (albeit requiring you to invest the labour somewhere else).
i believe that many hott researchers ended up working on internal hott might be an artifact of this.

>they had found a way to reasonably realize most of HoTT computably
this is less important than you might think, and mostly an artifact of type theorists being utterly autistic about using proofs by reflection.

>>12342128
Lmao replace tech nerd with EE, and you’ve got /sci/

>>12343896
>and it isn't obvious to me that the hott approach won't just be similarly laborious (albeit requiring you to invest the labour somewhere else).
I agree with this. The most popular direct implementation effort I know about for hott is lean, and lean fucking sucks all things considered

How does this help me code a fizzbuzz?

>>12344993
If you're using Haskell, you can write better fizzbuzz.

>>12344993
Clearly the I is fizz and the T is buzz
Bro you’re not very good at this

>>12344993
It allows you to implement fizzbuzz as pure function, with the Writer (WriterT) monad.
The real question is, how does this not help me code a fizzbuzz?

>>12344993
>>12345166
which category theory books are these? I have only read through some of Mac Clane's book and Hatcher's topology up to chapter 3

Asked this in the math thread but didn't get any replies, figured this might be a more relevant question for CSbros. Trying to get into Bayesian statistics and reading this book. It's really good but I was wondering if you guys know of any good youtube channels or other resources you would recommend.

>>12345249
Bayesian? More like Gayesian.

>>12313199
>>12313374
>>12314383
Ok Aaronson haters, convince me that there are good research-related reasons for me to hate the guy. His personality being obnoxiously "nerdy" or his political/whatever opinions being XYZ are not valid reasons.

>>12345252
I want to study network science and AI, fR*quentoid.

>>12344849
Erm, Lean pretty much abandoned the HoTT fad in favor of hardcore classical mathematics. The mathlib project embraces excluded middle and Hilbert's epsilon.

Most HoTT stuff I've heard about being actively developed is written in Coq or cubical Agda.

>>12345291
>Erm,

>>12336640
From a type-theoretic point of view, sheaves are pretty lame though. It's a crappy side-effect that relies on subtly non-constructive features to be actually any use. I know that Thierry has been working on that for great justice, but it's still not very convincing.

>>12345299
Care to elaborate, my fellow retarded onahole?

>>12345249
>youtube channels

You already have a great textbook on your hands. Why would you want a soitube video for?

>>12345260
>good research-related reasons for me to hate the guy
his contributions to physics are either trivial or overshadowed by his coauthors. One good paper with Susskind as the PI. The other was a meme paper about CTC's where he was the only computer scientist in the room...and he was swinging his dick around like he had researched more than they had.
>other reasons
researchers in CS like watrous are better mathematicians than aaronson. researchers in CS like gosset are better physicists than aaronson. researchers in CS like Arora are better computer scientists than he is.
The common link? None of them talk nearly as much shit nor constantly babble about "my personality trait is being a bullied nerd who never got over it," much less crying over twitter on anyone who doesn't like him.

The only good contribution from him I remember that's actually his is the BosonSampling paper, and that's about it.

>>12345346
I will reply to your post stripped of the inane garbage portions:

>his contributions to physics are either trivial or overshadowed by his coauthors. One good paper with Susskind as the PI. The other was a meme paper about CTC's where he was the only computer scientist in the room...

>researchers in CS like watrous are better mathematicians than aaronson. researchers in CS like gosset are better physicists than aaronson. researchers in CS like Arora are better computer scientists than he is.

>The only good contribution from him I remember that's actually his is the BosonSampling paper, and that's about it.

In regards to your complaints about his contributions in physics, I can't opine because it's not my field. Regardless, your complaints as a whole seem to be "he's not as good as X". Why is this a bad thing? For someone to be the best then everyone else has to be the not-best, and the not-best provide a valuable service by building the scaffolding the best will use to tackle the hardest problems. Besides this, his miscellaneous contributions like his blog posts about complexity theory or his maintenance of the complexity zoo are valuable too (and in some aspects more valuable than "hard" research since they publicise the field(s) to other people).
So yeah, as I suspected you hate the guy for superfluous reasons.

>>12345342
>see concept
>explanation is so/so and I've spent 30 straight minutes trying to understand
>could either ponder the idea for another 5 hours until I get it or just see someone else's explanation and potentially get it in the next 20 minutes
You gotta diversify your learning resources if you don't have a professor with office hours anon.

>>12345391
>"he's not as good as X". Why is this a bad thing? For someone to be the best then everyone else has to be the not-best, and the not-best provide a valuable service by building the scaffolding the best will use to tackle the hardest problems
perhaps I wasn't clear. I feel like his contributions don't and haven't done much to advance the field for a very long time, but he still addresses himself as a very big deal. I don't take interest in his side of quantum computing, and I feel like the most serious contributions to quantum information at large do not come from him - I'm more fond of Vidick's work than his own.
> Besides this, his miscellaneous contributions like his blog posts
contributions? you mean incel ranting despite having a wife? you mean constantly focusing in on his personal trauma? about trying to constantly score nerd cred? you mean always taking the bait zeilberger and motl put down, and responding with the most nothingburger "we did it reddit" responses? you mean contributing to the worst popsci parts of CS?
>complexity zoo
I like the complexity zoo, but not aaronson as teh guide through it. nobody but people who like complexity already actually give a shit about it
>superfluous reasons
no, I hate him because I don't think he produces good enough research to be interesting, yet he talks like he's the shit, AND he constantly interjects himself into the affairs of people doing better stuff because he has an obsession to be the face of theoretical computer science. He makes the research environment that much more uncomfortable.
Scott has a big mouth and nothing else

>>12345260
>>12345391
Lmao I didn't know Scott left reddit and twitter to start scouring 4chan for people who don't sing his praises. Gonna talk about how nerds are the most oppressed class of them all for the 10000000th? Or how "the rules of physics are SO coooOOol guys!"

>>12345391
Hey Scotty, i loved how buttmad you got about Motl.
Go hide in your basement.

>>12345391
>>12345260
>the state of computer """""scientists"""""" who like to pretend they're physicists
some of you are really good, but most of you are trash

>>12344849
what sucks so much about it? the things i can think of:
>it breaks type theoretical properties in favor of making life easier for classical mathematicians. this isn't much of an issue imo, since classical mathematicians won't expect these properties to hold anyways. for automation, you can always use meta-programming.
>the metaprogramming framework is slow. afaik the next version should be a lot faster, though.
>the current automation seems to be fairly hacky, possibly as a result of it being such a young project.
>you don't get all the bells and whistles of e.g. agda, but in turn lean doesn't have a proof of false every couple of months.


>>12345291
>The mathlib project embraces excluded middle and Hilbert's epsilon
this is not a problem in and of itself. you can do classical mathematics in hott! the issue, if you want to call it that, is that lean implements definitional proof irrelevance.
anon is probably referring to the lean2 hott library.

>>12345393
oof

>>12345266
just use the /sci/ wiki

>>12303692
It hasn't been for the last 7+ years.

Is there a language that is more mathematical than Haskell?

>>12331657
i am not scientist/resreacher, how ever in my opinion informatic concepts are near to death. for example theory of computation is not critical for information processing, but signal processing is essential for information processing. i cant understand how this field can not to be center of computer science. signals not need to be electrical. it can by study as information processes independly from type of information carrier, in same way as logic gate (elemental information processing system) can be realised physicaly using mechanical switches, electronical or even protein-based

>>12349921
scratch

>>12349921
Coq

>>12350050
>for example theory of computation is not critical for information processing,
>it's all sig proc
I don't think information is the center of CS since computation is a more fundamental idea, but your opinion is like entirely a naive EE undergrad's one to say that the theory of computation is not critical for information processing, especially when communication complexity and error correcting codes are incredibly popular topics with high impact.

>>12349921
APL or J

>>12349921
ALGOL

Which branches of maths are (most) important for Computer Science / Programming in general?

>>12351929
Abstract algebra, combinatorics, mathematical logic are core. Lots of other mathematics have become fundamental and relevant as time has went on - Fourier analysis is important to talk about probablistic characterizations of complexity classes via the PCP theorem, functional analysis for sketching algorithms, etc etc.

>>12352005
Why the fuck would you mention all that other shit but not category theory?

>>12352392
>category theory
There's not much math in category theory to begin with, let alone math that is relevant to CS. What little there is has essentially already been subsumed under logic.

You may be making the common mistake of assuming that anything capable of expressing math must be a branch of math. For example, the English language is capable of expressing math, but no one is turning to English majors for mathematical insight.

>>12352458
You're arguing that Category Theory is not a branch of Mathematics? Get the fuck out of /sci/.

>>12352458
>category theory not relevant to CS
What is Haskell?

>>12352590
>What is Haskell?
A programming language that exploits the vocabulary of category theory by extending the Curry-Howard correspondence with non-rigorous analogies.
http://math.andrej.com/2016/08/06/hask-is-not-a-category/
It can be considered an application of category theory like economics is an application of math, but if anyone claims to have written a Haskell program refuting a theorem of category theory, there is little doubt that the error will be traced back to Haskell's faulty implementation rather than that category theory is inconsistent.

>>12352392
>Why the fuck would you mention all that other shit but not category theory?
Category theory is getting increasingly important in CS, but I wouldn't say that it's quite as fundamental to learn at first, even if you want to get into PLT

I like the OP image in this thread very much. It is the most mathematical image I've seen in a while.

>>12353581
I wish when the real Tooker was on 4chan

>>12352509
>noooo, not my heckin pullback diagraminos, not the fricken left adjointerinos and the colimiterinos

How to get into complexity theory?

>>12353796
Computational complexity? Any decent book on algorithms and data structures will probably have the basics.

I NEED an infographic on a pathway of books to read to get familiar with CS like pic related. Ples help me out if anything is available otherwise one should be made.

>>12353807

>>12353807
That’s one of the most shitty lists I’ve ever seen in my life Christ.
And of course it has an anime girl on it saying some stupid bullshit.
Tranny spotted and most likely written by a transfag, explaining why it’s so fucking dumb.
I fucking hate them so god damn much, fuck transgenders. And they call everyone fascists and white supremicists.
Shit like that makes me actually want to oppress them, unironically, everytime they use the term fascist or white supremicist in the stupidest fucking way that adds to the counter of how many I’ll shoot down.

>>12353796
arora & barak

>>12353821
This seems nice for the codemonkey programming aspect, what about the mathematical theory aspects doe?

this thread wont die, huh

>>12353917
Of course not, it's incredibly based

CS isn't math. Abusing basic results from algebra and topology doesn't make you math, same as physics isn't math. The only "CS" that I would consider math is (non)linear optimization and numerical analysis.
Computation is not math
>>12303760
kek that is not math, and it disgusts me how you csfags distort the real math

>>12354317
Applied math is math, you dingus.

>>12354337
>Applied math is math
Wrong. (((applied math))) is an abomination

>>12354317
>>12354355
>relativization and Turing computability isn’t one of the purest parts of math
>areas like geometric complexity aren’t pure
Kill yourself

>>12354420
Anything tainted by CS is unpure by definition

>>12354431
>by definition
maybe if you did some more CS you'd have a better understanding of what this means

>>12353917
Two weeks

>>12354431
Not an argument.
What about the theory of expanders, used with thin groups for number theory? Most of the existing knowledge of expanders comes from CS departments, and I don’t think I’ve ever heard anybody say modern number theory is unpure.

>>12354317
>Abusing basic results from algebra
such as? you're telling me nullstellansatz used in arithmetic combinatorics is somehow an "abuse" of "basic" results?
>topology
such as? How is using the adjunction between the realization and singular set functors in UMAP an abuse of topology? How is it basic?
>doesn't make you math
Ah, but operator theory and differential geometry has had extensive pure study in physics departments - the same goes for combinatorics and algebra in the CS departments. Either way, you're full of shit
>muh nonlinear optimization and numerical analysis
confirmed for not know what CS is or what is studies. Don't talk about pure math unless you know what finite injury or topological games are, freshie undergrad.

>>12354941
>>12355295
Dilate and never come near me on my math ever again. And wash your programming socks.

>>12350050
>electrical engineer thinks anything computers is but a subset of his capabilities, and anything that doesn't fit in his scheme is superfluous shit "near to death"
Look, I won't talk about your motors, because I have absolutely no clue about them, but please return the favour and do not talk about shit you don't understand.

>>12355309
>assmad someone called him out on his undergrad
i don't give a flying fuck about sorting algorithms, but I know that if someone came at you with some way Noether Normalization was used in sorting lower bounds, you'd call it trivial, highschooler
>MY math
why do brainlets constantly try to claim for themselves shit they never did?

>>12303630
I want to fuck that thing. why?

Can someone point me to where I might learn more about the language mentioned in this answer? Any papers or books that go over the algebra of this calculus? It doesn't seem like standard lambda calculus.
https://mathoverflow.net/questions/11540/what-are-the-most-attractive-turing-undecidable-problems-in-mathematics/25926#25926

>>12356629
>https://mathoverflow.net/questions/11540/what-are-the-most-attractive-turing-undecidable-problems-in-mathematics/25926#25926
it's a self explanatory post. but you might be able to get some basic explanation for the *flavor* of his post from a book like Sipser

>>12356018
you didn't prove that

-obama

>>12356691
yes i did

- nixon

>>12356608
you tell me

>>12356608
>t. F42

>>12354317
>kek that is not math
Analytic combinatorics is nothing but math

bump

>>12303760
would anyone be interested in a reading group for this book in the future

>>12306411
Underrated post

>>12358636
bump

How would I go about learning cryptography? Bitcoin exposed me to things like Zk-Snarks and public key cryptography and I'd like to understand the math behind it.

>>12361652
Visit >>>/g/sec where this came up recently:
>There are some questions now and then about crypto. Here is a course from 1944:
https://radionerds.com/images/b/bb/TM_11-485.pdf

>>12361652
There are two parts of cryptograph: theory and engineering. To learn theory, you should learn abstract algebra and number theory. You should do all of the group and ring theory in Artin, but just the basics of groups and rings will be enough to start reading cryptography literature. You SHOULD go through all of the algebra and make sure your fundamentals are there - then you can pick up what you need at the pace you want. There aren't very good number theory textbooks out for undergrad, but this book:
https://www.math.brown.edu/johsilve/MathCryptoHome.html
should suffice.

As for engineering a cryptosystem, these books are great introductions:
https://www.schneier.com/books/cryptography-engineering
http://cacr.uwaterloo.ca/hac/

>>12361652
>Bitcoin exposed me to things like Zk-Snarks and public key cryptography
lmao @ /biz/

>>12303692
That's like saying /tv/ is the television and film board

>>12318272
>>12318593
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004295#pcbi.1004295.s010

Some interesting ML/DL papers and examples with quick summaries

https://www.youtube.com/watch?v=INBjNZsnfy4
>Previously, state of the art non-DL pose estimation (think of it as modeling objects in 3d space given a single perspective 2d image/video of it) relied mostly on human intervention to define "keypoints" to model the motion of objects, which are important patches of the image somehow associated with the object (ex. corners) which the model can track and accordingly estimate location, orientation, motion. This obviously is bad if you want a fully automated computer vision pipeline. These guys introduce a new method of estimating keypoints without supervision.

https://openreview.net/forum?id=Bygh9j09KX
>DL computer vision methods extract "texture" features as opposed to "shape" features like sillhouette/edges. In contrast, humans can consider both texture and shape features. This leads to poor computer vision performance when the texture is changed or removed compared to humans. It's theorized that L-infinity bounded adversarial attacks make use of this lack of texture robustness.

obligatory neat video: https://www.youtube.com/watch?v=MIbFvK2S9g8

>>12363372
To be fair, I have seen some tv and movie threads in between the cunny threads. They're not common, but they do exist.

How do I get into CS research in a 3rd world shithole? The research here (even in math) is very meme-y.

>>12363417
Nice
>>12365260
Try and contact professors from nearby countries?

>>12363372
Or that /gif/ is a porn board and not a larping race war

Induction is fake, prove me wrong

>>12366591
>my model is wrong, therefore every model is fake
nice brain

>>12366691
i don't see what the problem is here ;^)

>>12367026
>>12366591
Inductive arguments can be valid but in your case the *harkle harkle* theorem is not sound albeit valid, QED. Any more homework questions?

>>12314383
>TT autists are too busy playing around to shitpost
This. I'm having fun with types-as-closures lately. Making everything into an equational theory on top of the untyped lambda calculus is neat as fuck. And the statistical reasoning on top of it is just icing on the cake.
http://www.math.cmu.edu/~fho/notes/2007/definable_closures.pdf
http://www.math.cmu.edu/~fho/notes/2007/definable_closures.pdf
http://www.math.cmu.edu/~fho/fritz-thesis-draft.pdf

>>12367509
i always want to learn more about type theory. should I start from the HoTT book? I have algebra, analysis, and topology (point set and algebraic) underneath my belt.

>>12367618
It's not too hard of a read since it is written in a very self-contained way. If you want a good foundation on classical type theory as well, look into Pierce's books.

Algebra is probably the most relevant, but algebraic topology will be helpful with a deeper understanding of HoTT semantics.

>>12367632
>Pierce's books
got it. Question:
there's been a lot of stuff here posted about recursion theory, about type theory, and geometric complexity. I like a good amount of these topics, but I always notice there's a split between math and CS faculty studying them, with bias being towards math faculty. Would you say there are more mathematicians investigating these, or that I'd be at a disadvantage if I went to a CS PhD program with the intent to study these rather than a math PhD? I have honors undergrad / early undergrad proficiency in both and research with professors done in both departments.

>>12367662
>early undergrad
meant to say early grad. So for CS, I'm past Arora-Barak in complexity, and for math I'm around Reed-Simon in functional analysis.

>>12367662
As it gets closer to logic and recursion theory, the split between math an CS get fuzzy. It really depends on what you are after. A good CS program will have a good amount of math, but its going to be heavily tilted to the algebraic side.

If you are interested in foundations, theory of computation, complexity, logic, or the likethen CS will probably be what you are looking for, although pure logic will certainly have its fair share of pure mathematicians researcher.

If you prefer the aspects of type theory that relate to AT or Homotopy theory, then you will probably want to do a math PhD.,

>>12367828
That makes sense. I guess what I'm also asking is that neither a math program nor a CS program for the PhD would stop me from studying stuff 'closer' to the other subject in the future, right? I just don't want to pigeonhole myself is all.
I know there are a good amount of CS PhD's who have math appointments, like Srivastava and Grochow, but I'm still kind of nervous.

>>12367953
I'd definitely look into what research programs your school has. If there's CS research that's closer, go for that.
Math is a lot more broad, it really depends on what you have access to. Your advisor will probably be able to help find what programs closest align to your research interest.
But if everybody is working on differential equations and analysis, don't expect much in the way of CS topics, expect for when it's relating to numerical analysis.

>>12367953
Also, the thing with CS, is that you will be picking up a more 'eclectic' set of mathematics. Your research path might take you to math that you didn't expect to need, or even knew existed.
A math PhD, might take you into more mainstream mathematics. CS topics tend to reach far and wide across the spectrum and bring together a lot of seemingly unrelated fields.

HoTT being a motivating example here, another being linear logic and its relations to functional analysis

>>12367997
>>12368019
Hm, but I've noticed that there are a lot of math researchers in CS who do "traditional mathematics" more than CS researchers - I have noticed that CS researchers have unorthodox toolsets (fourier analysis alongside combinatorics comes to mind), but would you say that going for the CS PhD would "ruin" a good math background, or that the math PhD has a natural advantage? I suspect the answer to my question is no, given the existence of Spielman's students, Baez's students, etc etc., but I'm just trying to collect wisdom and experiences here since I'm applying next fall.

Thanks again for all the advice!

>>12368066
All I can give you is my experience, but I think I am not really lacking any mathematical maturity from the path I took.
I have more of the unorthodox CS background, I recently picked up Cohen's Number Theory for some crypto research, and didn't feel like I was missing any background knowledge that I hadn't encountered.
I'm sure there is likely depth in some topics that I am missing, probably on the analytic side of things, but in general I am able to follow along well in more orthodox pure math texts.

>>12368113
That's fair. I've heard the same sentiment from most CS researchers I've encountered: you pick up what you need with ease, but CS researchers *tend* to stick to discrete phenomena more as a habit of culture rather than need or inability to do the alternative. I'm personally interested in places where algebra and topology show up in CS theory - I feel like there's a lot of fruitful results / theory that has yet to be discovered here that is way past the seminal yet basic stuff from the 70s. But I guess I got nervous when I saw most of the big names (Kleene, Cook, Spielman, etc.) had math PhDs, but that's more a function of the fact that the CS PhD didn't really *exist* until well into these giants' career.

So in the end, I guess I should look more for advisors and departments that do what I like rather than looking at the title of the PhD (except maybe to get an idea of the requirements / quals)?

>>12368145
Yep. The best thing you can do is to shop around to see where you best fit in
There's a lot of potential with topological reasoning in CS. Fundamental groups and their relation to looping paths of inside the spaces of rewrite systems has been something I've been preparing to work on. There's a lot of room to make serious research contributions in what I have seen from your interests.

>>12368177
> lot of room to make serious research contributions in what I have seen from your interests.
phew. I'm interested in the stuff I mentioned before and parts of quantum information, particularly hamiltonian complexity as well. I think what's been giving me pause lately is that I've been trying to balance my interests between mathematical physics and CS, and it seemed like the intersection was empty or at least barren. I'm happy that these routes are (hopefully) fruitful.

bump>limit

>>12369531
That was reached about 11 hours before your post.

>>12369572
ok but i just woke up where is nuthread?

>>12369579
What stops you from creating it yourself?

>>12369588
meh
need to shower and eat now anyway buhbye