/sci/ - Science & Math » Thread #12893521

129KiB, 1009x614, CV_watching.jpg

View Same Google iqdb SauceNAO

/mcs/ - Mathematical CS General

Anonymous Sun 28 Mar 11:13:58 2021 No.12893521 View Reply Original Report

Quoted By: >>12894175 >>12899322 >>12900608 >>12905112 >>12905910 >>12925411 >>12934619 >>12943426 >>12946532 >>12949237 >>12951096 >>12956585 >>12962659 >>12967256

This thread is dedicated to the mathematical aspects of computation. It doesn't matter whether or not it's theoretical CS or applied. Old thread here: >>12841837

This thread features computer vision (CV).

Here are some resources for beginners:
https://teachyourselfcs.com/

Free books:
https://z-lib.org/
http://gen.lib.rus.ec/
https://spoon.wiki/Books

Here are some resources to get you started.

Linear algebra (free online!):
https://www.math.brown.edu/streil/papers/LADW/LADW.html
You only need to understand material from the first 4 chapters to get started with CV.

Computer Vision: Algorithms and Applications (free online!):
https://szeliski.org/Book/
The bottom of the page has multiple courses and slide sets on the same material, provided both by this instructor and many others.

Stanford CS 131 CV class (2018):
http://vision.stanford.edu/teaching/cs131_fall1819/syllabus.html
CS 131 (2014):
http://vision.stanford.edu/teaching/cs131_fall1415/index.html

Stanford 231 CV (harder than 131) lecture videos:
https://www.youtube.com/playlist?list=PLf7L7Kg8_FNxHATtLwDceyh72QQL9pvpQ

Anonymous

Anonymous Sun 28 Mar 2021 11:46:34 No.12893566 Report

Quoted By: >>12894527

Anyone know any good lecture videos for Computability Theory?

Anonymous

Anonymous Sun 28 Mar 2021 12:21:54 No.12893623 Report

Quoted By: >>12893629 >>12893922 >>12897393 >>12897763 >>12901084

lol, coders are not scientists and do not deserve an education. computer science is the new assembly line work.literally blue collar but monkeys believe they are white collars. They get the title of engineers to get a high dose vapid power trip and tell their parents they are not failure, while still having a shitty pay. LOL engineer my ass.

this is why all the big corps like apple push the state to make coding mandatory in public school, in order to increase competition among the monkeys and to avoid spending money on training them

Anonymous

Anonymous Sun 28 Mar 2021 12:23:30 No.12893629 Report

Quoted By:

>>12893623
'Coding' as you call it is not computer science, that's just programming. You don't need to be a computer scientist to do that.

Anonymous

Anonymous Sun 28 Mar 2021 14:30:56 No.12893922 Report

Quoted By: >>12894967 >>12951105

>>12893623
If you think computer scientists get the title of "engineer", then they are not computer scientists, but software engineers. The similarity between them is equal to car designers and mechanics.

Anonymous

Anonymous Sun 28 Mar 2021 14:45:21 No.12893958 Report

Quoted By: >>12894004 >>12895901

Any resource on integer linear programming?

Anonymous

Anonymous Sun 28 Mar 2021 15:05:21 No.12894004 Report

Quoted By: >>12894021

>>12893958
Schrijver is the standard intro

Anonymous

Anonymous Sun 28 Mar 2021 15:13:43 No.12894021 Report

Quoted By:

>>12894004
Thanks. I will check it out.

Anonymous

Anonymous Sun 28 Mar 2021 16:05:43 No.12894175 Report

Quoted By: >>12894620 >>12897232

>>12893521
I missed the poll in the last thread, but would be interested in opening up discussion for
>Quantum Information and Computing
>Programming Language Theory and Compilers
as well as cryptography / cryptographic algorithms. I am just an undergrad trying to find a niche, but i’m especially drawn towards the later. i’m also double majoring in math, and was hoping to get an opinion on what I should prioritize over the next two years for graduate school.

Anonymous

Anonymous Sun 28 Mar 2021 17:40:06 No.12894527 Report

Quoted By:

>>12893566
Use this:
https://www.youtube.com/playlist?list=PLbMVogVj5nJSd25WnSU144ZyGmsqjuKr3
Unironically one of the best set of recorded video lectures following Sipser’s pedagogy. You might want to watch on 1.25x if the prof is speaking too slowly, but otherwise his presentation is very clear and thorough, accent non intrusive, etc.. Unfortunately, the lectures only go up to the basics of Turing computability.

Anonymous

Anonymous Sun 28 Mar 2021 18:08:15 No.12894620 Report

Quoted By: >>12898286

>>12894175
If you were in the other thread, then you might remember that there’s an anon who posted who does post-quantum crypto. I was waiting for his gitlab to go up before dedicating a thread to cryptography, but I can definitely enqueue it as the next topic if there’s sufficient interest.
I’m no expert in cryptography, but I’ve done my fair share of classes and cursory research into lattice crypto. There’s no real upper bound on what is relevant, but there are some common foundations. I’m assuming you are comfortable with standard math/CS requirements like linear algebra. For math, I think you should probably prioritize:
ring theory (most of a second course in an abstract algebra full year sequence)

probability theory at least at the level of Sheldon Ross’s book (so just take it in the math dept)

combinatorics (Stanley is good but might be overkill. Graduate combinatorics has always served me well though)

information theory (cover-Thomas is standard, may not be standard ugrad course in your math/CS dept)

Elementary number theory (I think with enough algebra, you’ll be able to pick it up better than a full ugrad course. I’d still take it to be on the safe side)

For CS, of course you should understand:
Algorithmic design and analysis

Computability and complexity

And of course, the Handbook of Applied Cryptography is your best friend. It has a small primer on most topics above.
http://cacr.uwaterloo.ca/hac/

Anonymous

Anonymous Sun 28 Mar 2021 19:24:40 No.12894967 Report

Quoted By:

>>12893922
Sort of, except it’s more the case there many software engineers don’t deserve the title of engineer. That being said, there ARE a decent amount of professions in software that are actual engineering. I would be hard pressed to argue that cryptosystems people and the software team at JPL aren’t engineers doing engineering.

Anonymous

View Same Google iqdb SauceNAO AB5B937B-FD2D-47DD-9CD6-EC5BA6FC (...).png, 122KiB, 512x285

Anonymous Sun 28 Mar 2021 21:54:43 No.12895533 Report

Quoted By:

Here are some courses from CM, both of which use the CV book in the OP.

Computer vision:
http://www.cs.cmu.edu/~16385/

Computational photography:
http://graphics.cs.cmu.edu/courses/15-463/

Anonymous

Anonymous Sun 28 Mar 2021 22:43:58 No.12895712 Report

Quoted By: >>12898127

Oh, I forgot to mention. Here are some resources for OpenCV. a standard CV library used extensively in both academic and industrial contexts:
https://github.com/spmallick/learnopencv

Anonymous

Anonymous Sun 28 Mar 2021 23:21:29 No.12895901 Report

Quoted By: >>12896327

>>12893958
The bertsimas book is the best LP book, its used mainly for intro grad level courses and it is approachable with just a basic knowledge of proofs and linear algebra skills. The schrijver book is okay.... schrijver writes his proofs tersely so it can be hard to learn from at times, it also has typos. I recommend getting a separate text for IP, the most up to date one is the springer integer programming by cornejoul et all.

Anonymous

Anonymous Mon 29 Mar 2021 02:10:34 No.12896327 Report

Quoted By: >>12898125

>>12895901
Thanks. Any recommendation on combinatorial optimization as well?
I'm just reading Michel Goemans' lecture notes at the moment but I want a more thorough textbook.

Anonymous

Anonymous Mon 29 Mar 2021 08:33:54 No.12897232 Report

Quoted By: >>12898286

>>12894175
Are you a freshman?

Anonymous

Anonymous Mon 29 Mar 2021 10:12:59 No.12897393 Report

Quoted By:

>>12893623
Weak bait.

Anonymous

View Same Google iqdb SauceNAO wipe.gif, 98KiB, 220x150

Anonymous Mon 29 Mar 2021 12:39:53 No.12897763 Report

Quoted By:

>>12893623
oh no i am only a coder who works on blue collar work :(

Anonymous

Anonymous Mon 29 Mar 2021 15:00:55 No.12898125 Report

Quoted By:

>>12896327
Despite being dated I really love the rigour in Papademetriou and Steiglitz I think it helps you a lot when you think about developing new algorithms etc

Anonymous

Anonymous Mon 29 Mar 2021 15:01:17 No.12898127 Report

Quoted By: >>12898170 >>12948207

>>12895712
Ayyyy. I work in computer vision and this is basically my lifeblood. It's a pain in the ass to work properly with cudnn and cuda if you want any real-time kind of implementation.

Anonymous

Anonymous Mon 29 Mar 2021 15:18:02 No.12898170 Report

Quoted By:

>>12898127
oh nice. what's your work on, and how long have you been in the field?

Anonymous

Anonymous Mon 29 Mar 2021 15:46:31 No.12898286 Report

Quoted By: >>12898482

>>12894620
thank you. great information here to keep in mind going forward. I had a lot of the math on my radar, (taking probability theory next semester and combinatorics now), but hadn’t considered information theory. the only information theory course offered at my uni is graduate level with a couple of prereqs, but perhaps I can find a way to take it if it’s critical.
>>12897232
I am not, why do you ask?

Anonymous

Anonymous Mon 29 Mar 2021 16:25:52 No.12898482 Report

Quoted By:

>>12898286
>the only information theory course offered at my uni is graduate level with a couple of prereqs, but perhaps I can find a way to take it if it’s critical.
I wouldn't worry about it too much. Most of the information theory you need at your level is developed in the Handbook, and you'll be able to pick up what you need until graduate school by reading texts that have short primers on it.
While not all the way relevant to cryptography, you might find this useful to go through:
http://www.tcs.tifr.res.in/~jaikumar/Papers/EntropyAndCounting.pdf

Anonymous

Anonymous Mon 29 Mar 2021 17:15:32 No.12898645 Report

Quoted By: >>12898753 >>12898798 >>12958790

The state of ML research:
https://www.reddit.com/r/MachineLearning/comments/mdpb6v/d_dilemma_mathematically_wrong_icml_submission/
If there is going to be another AI bubble burst it will be because of things like these.
I know a lot of people in other fields (chem, bio, med) who use ML as a blackbox and they actually design their experiments incorrectly, but still got their work published anyway.
This whole thing is crazy.
Computer vision is still going strong though.

Anonymous

Anonymous Mon 29 Mar 2021 17:43:26 No.12898753 Report

Quoted By: >>12898930

>>12898645
holy shit, that's fucked. Is this within ML theory or among people in practice? I always got the impression that people in ML theory were a bit more careful about their work.
>vision is still going strong
I think CV is still one of the killer apps of learning techniques and has had so much attention from all sides of STEM that it feels one of the most "honest" subsets of learning-based research

Anonymous

Anonymous Mon 29 Mar 2021 17:54:46 No.12898798 Report

Quoted By: >>12898811

>>12898645
it's the same shit in every field that even remotely touches $$$
read Leonid Schneider for a cute summary of this in bio*
even better is the reaction of "the authorities" to scientific misconduct - "hey, photoshop for quickly conjuring up fake data to support our papers is one of the basic skills every biochem researcher should master"
academia is dead

Anonymous

Anonymous Mon 29 Mar 2021 17:56:43 No.12898811 Report

Quoted By: >>12898992

>>12898798
idk, I'm not aware that this sort of shit happening in the quantum information community or even just theoretical CS

Anonymous

Anonymous Mon 29 Mar 2021 18:25:23 No.12898930 Report

Quoted By:

>>12898753
>I always got the impression that people in ML theory were a bit more careful about their work.
It really depends on the PI honestly. I know some advisors who are really careful with the proofs, almost perfectionism level of carefulness.
Others pretty much leave the heavy proof-check to PhD students and there will likely be mistakes. Some mistakes get caught in the review process, some don't.

Anonymous

Anonymous Mon 29 Mar 2021 18:40:37 No.12898992 Report

Quoted By: >>12899219 >>12900501

>>12898811
>quantum information community
give it a decade or two, when quantum memes are sufficiently real and usable for generating $$$
>theoretical CS
too broad
i presume you'd agree that someone exploring "connections between CSP and conjugacy classes of groups of automorphisms of finite abelian groups" might not be financially motivated, correct?

Anonymous

Anonymous Mon 29 Mar 2021 18:54:47 No.12899056 Report

Quoted By: >>12899256

Look at all of the retards in this thread:
>>12889728
Even so, /sci/ will continue to make fun of CS.

Anonymous

Anonymous Mon 29 Mar 2021 19:41:43 No.12899219 Report

Quoted By:

>>12898992
>give it a decade or two, when quantum memes are sufficiently real and usable for generating $$$
I mean, NISQ is here and they can be used for pretty important things now
>exploring "connections between CSP and conjugacy classes of groups of automorphisms of finite abelian groups" might not be financially motivated, correct?
not exactly, but at the same time, I don't see this from the subset of TCS researchers in, say, cryptography, where the money in the field is huge as well. Then again, cryptography is probably too mission critical to accept failure

Anonymous

Anonymous Mon 29 Mar 2021 19:48:05 No.12899256 Report

Quoted By: >>12903696 >>12903727

>>12899056
CS is a relatively young field that's somewhere between pure math and engineering. The low hanging fruit is still being phased out of systems, but they've been eradicated from the theoretical side just about entirely now. It's unfortunately going to get bullied for the next few decades until it becomes incredibly prestigious when newer generations lose their hangups and embrace the progress and study with arms open.
We saw this with aerospace engineering. We saw this with combinatorics (hell, people still joke about combinatorics not being math despite it being one of the hardest subfields with lots of mature resutls), we saw this with bioinformatics and bioengineering, and I'm confident we're gonna see this with computer science.

Side note: someone mentioned this on /g/, and I think it describes so much of /sci/ perfectly:
"Computer science is a big deal, and that usually pisses everyone off in traditional fields of math and engineering because they’ve conditioned themselves to think they’re smarter since CS is the new kid on the block. EE students in particular tend to be the most assmad since they see themselves as the “owners” of computing technology and all the clout and prestige it generates."

Anonymous

Anonymous Mon 29 Mar 2021 20:04:27 No.12899322 Report

Quoted By: >>12899351 >>12899360

>>12893521
Why would you need math for counter strike?

Anonymous

Anonymous Mon 29 Mar 2021 20:12:22 No.12899351 Report

Quoted By: >>12899360 >>12904209

>>12899322
Someone please create a game where terrorists and police battle each other using (high school) math.

Anonymous

Anonymous Mon 29 Mar 2021 20:15:13 No.12899360 Report

Quoted By: >>12899365

>>12899322
kek
>>12899351
why highschool?

Anonymous

Anonymous Mon 29 Mar 2021 20:16:44 No.12899365 Report

Quoted By: >>12899603

>>12899360
Mass appeals.

Anonymous

Anonymous Mon 29 Mar 2021 21:30:41 No.12899603 Report

Quoted By:

>>12899365
>he doesn't do triple barnett integrals in his head
baka

Anonymous

Anonymous Tue 30 Mar 2021 01:31:15 No.12900501 Report

Quoted By:

>>12898992
> when quantum memes are sufficiently real and usable for generating $$$
we're already seeing this on the business side.

Anonymous

Anonymous Tue 30 Mar 2021 02:08:12 No.12900608 Report

Quoted By: >>12900633 >>12900739

>>12893521
anon, how often do you touch computer architecture in your work / research? my uni offers one of the top computer architecture courses / tracks in the country. however, making room for it in my schedule would require me to take less math, and i’m slightly more interested in theory than systems.

Anonymous

Anonymous Tue 30 Mar 2021 02:17:02 No.12900633 Report

Quoted By: >>12900739

>>12900608
You don't need it, leave it to the computer engineers. While it can in some cases be good to broaden your horizons, this won't really help. Computer science has models of computation and they are sufficient for the purposes of doing computer science.

Anonymous

Anonymous Tue 30 Mar 2021 02:20:26 No.12900644 Report

Quoted By: >>12900740

can CS be used to compute figure drawing?

Anonymous

View Same Google iqdb SauceNAO D08A3936-F8B8-49A4-8A17-3031C055 (...).jpg, 325KiB, 918x960

Anonymous Tue 30 Mar 2021 02:24:43 No.12900657 Report

Quoted By: >>12900677 >>12936415

Just started SICP today, I’m at 1.1.7. Weirdly my philosophy degree is coming in handy since I studied symbolic and modal logic in uni. Frankly the best guys in my logic classes were CS masters students

Anonymous

Anonymous Tue 30 Mar 2021 02:32:03 No.12900677 Report

Quoted By: >>12900795

>>12900657
coincidentally I am taking a modal logic course right now. what are some of the most interesting applications you have seen?

Anonymous

Anonymous Tue 30 Mar 2021 02:55:53 No.12900739 Report

Quoted By: >>12900757

>>12900608
>anon, how often do you touch computer architecture in your work / research?
My own research? I personally work on stuff related to hardware in the quantum setting, between photonics on the hardware design side and toric codes / ECC on the "making it fault tolerant" side.
I have a friend who's in systems CS, and he works a lot on memory virtualization and TLB research, which is usually applied to chip designs.
> require me to take less math, and i’m slightly more interested in theory than systems.
follow your interests. You could always ask the professor if you can audit the class to get the material / experience without cutting into your current professional interests.
>>12900633
>You don't need it, leave it to the computer engineers
this is shitty advice lmao. Every good CS program should have good comp arch and systems / OS. Regardless, to address the rest
> Computer science has models of computation and they are sufficient for the purposes of doing computer science.
This is a reductive take because computer science shows up at the low level / chip design anyway. There's what I mentioned above, which usually affects cache and virtualization / hardware address space aliasing and disk policy, but there's also stuff like studying how to queue instructions efficiently by fetching multiple and analyzing dependency DAGs in order to parallelize, using forms of deep learning to schedule well, etc etc. CS absolutely has relevance to architecture design.

Anonymous

Anonymous Tue 30 Mar 2021 02:56:56 No.12900740 Report

Quoted By: >>12900751

>>12900644
what does compute figure drawing mean in this context?

Anonymous

Anonymous Tue 30 Mar 2021 03:00:46 No.12900751 Report

Quoted By:

>>12900740
compute a drawing. so that it looks correct.

Anonymous

Anonymous Tue 30 Mar 2021 03:03:08 No.12900757 Report

Quoted By: >>12900801

>>12900739
Your example sounds more like computer engineering to begin with, when we're in a 'mathematical computer science' general. I still don't think a theoretical computer scientist would benefit from a very technical computer engineering course.

Anonymous

Anonymous Tue 30 Mar 2021 03:13:50 No.12900795 Report

Quoted By:

>>12900677
Well I’m about 10 pages in so far so while programming resembles cp logic I haven’t reached anything modal yet. I remember the CS guys being interested in system completeness. One guy said there was a connection to machine learning but I’m not informed about that. My prof also really emphasized the semantic tableaux which is such a useful tool. Are you from a phil or compsci background? I’d love to hear some insight or connections between the two fields if you’ve got any

Anonymous

Anonymous Tue 30 Mar 2021 03:14:47 No.12900801 Report

Quoted By: >>12900821

>>12900757
>Your example sounds more like computer engineering to begin with
I mean these are literally common topics you'd learn in any CS class for OS, parallel computing, or advanced computer architecture.
>when we're in a 'mathematical computer science' general
while that's true, anon was asking about his general CS degree / track.
>I still don't think a theoretical computer scientist would benefit from a very technical computer engineering course.
Ehhh, hard disagree. There are interesting theory problems that show up in the systems setting that have computer engineering applications but are TCS problems nonetheless. It won't teach you theory, and I wouldn't suggest taking these comp arch courses *over* theory courses if you must make a choice, but as a theorist, I've found that engineering subfields of CS and the applied stuff is usually a rich bed for theory.
To name a few examples, we have Lamport's memetic Part Time Parliament paper on the consensus in systems: https://lamport.azurewebsites.net/pubs/paxos-simple.pdf
Another would be the Godel Prize winner, the asynchronous computability theorem for distributed systems: http://people.csail.mit.edu/shanir/publications/HS-topology.pdf
Then there's a huge subset of algorithms made to study problems in cache-obliviousness, online algorithms, etc etc. which all have relevance in memory. In particular, the study of B^epsilon-trees in pure algorithms has large motivation in write optimized I/O: http://www.betrfs.org/
So your point is tantamount to saying, "I don't think a combinatorialist would benefit from a topology class or an analysis class" when various parts of algebraic topology are studied using combinatorics and analytic combinatorics exists. Not all of these examples touch chip design, though a theorist could break into VLSI by studying the relevant combinatorial optimization / LP's involved, but the point is that theory is a *way of thinking* about problems rather than just a body of knowledge.

Anonymous

Anonymous Tue 30 Mar 2021 03:21:39 No.12900821 Report

Quoted By:

>>12900801
You make a fair point, I agree that a lot of great research comes from people who are not just narrow minded specialists but have broader horizons.
I still think if the course in particular is aimed at engineers it might contain a lot of detail that's not entirely useful, but I don't know.
The suggestion to audit it is best I think. Or just sit in if the professor allows it (they usually do).

Anonymous

Anonymous Tue 30 Mar 2021 03:43:33 No.12900895 Report

Quoted By: >>12902984 >>12906751

can anybody tell me if this means anything? h(t) ??(t)>
It's from the ramblings of a schizo I expect it means absolutely nothing but i'm curious nevertheless.

Anonymous

Anonymous Tue 30 Mar 2021 05:10:30 No.12901084 Report

Quoted By: >>12901108

>>12893623
Nice bait. I would say (roughly speaking)Comp Sci is to coding what Pure math is to engineering. This analogy falls apart under too much scrutiny, but yeah.

Anonymous

Anonymous Tue 30 Mar 2021 05:20:39 No.12901108 Report

Quoted By:

>>12901084
sorta but not really, since the whole idea behind computer science is computability / constructability. So there is definitely room for software systems to be a part of CS, since the whole idea is that we can actually implement a lot of the results, but unfortunately, a lot of industry software positions are less about engineering and more about development or UX.

Anonymous

Anonymous Tue 30 Mar 2021 06:45:11 No.12901312 Report

Quoted By:

Any algorithmic information chads out here?
http://www.scholarpedia.org/article/Algorithmic_information_theory

Anonymous

Anonymous Tue 30 Mar 2021 14:09:01 No.12902984 Report

Quoted By:

>>12900895
Might have to do with entropy

Anonymous

Anonymous Tue 30 Mar 2021 16:02:33 No.12903696 Report

Quoted By: >>12903727 >>12903845 >>12905578

>>12899256
I think the biggest factor is it just isn’t rigorous at most institutions, so most Computer Science degrees are really more of a Software Engineering degree. There’s a lot of respect for grad/post-grad Computer Science.

Anonymous

Anonymous Tue 30 Mar 2021 16:06:34 No.12903727 Report

Quoted By: >>12903845

>>12899256
>>12903696
Didn't some computer scientist win the Abel prize recently?

Anonymous

Anonymous Tue 30 Mar 2021 16:28:24 No.12903845 Report

Quoted By:

>>12903696
>There’s a lot of respect for grad/post-grad Computer Science.
I still see some researchers in some parts of math and some parts of physics scoff at CS thinking it's just the theory of simulating stuff in math and physics. But I suppose that's no different than math researchers who scoff at physics for being applied differential geometry. Stupid people will be stupid.
>>12903727
I think Lovasz (mathematician in combinatorics who has a lot of CS adjacent work) and Wigderson (computer scientist / complexity theorist who started out decades ago in math but now exclusively does complexity theory) jointly won the Abel Prize. I think the prize going to two people associated heavily with computer science is important in the long term for CS being seen more favorably among people in other fields, even if the prestige of the prize is at most artificial. I think the PR it does for the field among the traditional mathematicians is worth its weight in gold.

I'd rather the prize be used so that results in CS are mainstays for results in math and vice versa, in the same way math, statistics, and physics have done for years

Anonymous

Anonymous Tue 30 Mar 2021 17:25:41 No.12904209 Report

Quoted By:

>>12899351
"Fuck guys, they have SICP smurf"

Anonymous

View Same Google iqdb SauceNAO geometric_methods_in_optimizatio (...).png, 257KiB, 520x240

Anonymous Tue 30 Mar 2021 18:56:20 No.12904523 Report

Quoted By: >>12905588

https://simons.berkeley.edu/programs/gmos2021
>The geometric toolbox that underlies optimization and sampling was initiated in the study of partial differential equations (PDEs) and has evolved into different mathematical disciplines: probability, calculus of variations, analysis and geometry.
>A central goal of this program is to develop and promote a geometric approach to various computational problems in sampling, optimization, and PDEs. For example, the geometry of Optimal Transport has been instrumental to establish fruitful connections between diffusion processes, gradient flows, and diffusive PDEs by eliciting hidden convexity.
HYPE
Maybe we can get more applied analysisfags and engineering students into CS, or at least this thread.

Anonymous

Anonymous Tue 30 Mar 2021 21:03:23 No.12905112 Report

Quoted By: >>12905144

>>12893521
Learning about Neural Networks and stuff and how a NN can be seen as a function approximator. Would it be possible for a NN to generate a analytic function or NN for a 3d model?
Something like describing the model of the face of a person as a series of ellipses or the combination of them.
Could such a NN improve rendering times or help somehow else?

Anonymous

Anonymous Tue 30 Mar 2021 21:11:34 No.12905144 Report

Quoted By:

>>12905112
They could but they can be noisy. Here's some related work that generate 3D triangular mesh based on 2D images.
https://openaccess.thecvf.com/content_ECCV_2018/papers/Nanyang_Wang_Pixel2Mesh_Generating_3D_ECCV_2018_paper.pdf
If you can think of a good objective function / regularization that makes smooth model then you are going to be very popular in the field.

Anonymous

Anonymous Tue 30 Mar 2021 23:17:30 No.12905578 Report

Quoted By:

>>12903696
> it just isn’t rigorous at most institutions, so most Computer Science degrees are really more of a Software Engineering degree.
lel a proper soft eng degree would be some form of engineering.
the real solution is to have CS become CSE everywhere, where the classes are harder and you have a harsher STEM requirement. Hell even put it in the school of engineering.

Anonymous

Anonymous Tue 30 Mar 2021 23:19:59 No.12905588 Report

Quoted By: >>12905653 >>12906450

>>12904523
but continuous optimization is the gay optimization.

>t. discrete optimizer

Anonymous

Anonymous Tue 30 Mar 2021 23:48:40 No.12905653 Report

Quoted By:

>>12905588
embrace the smooth (brain)

Anonymous

Anonymous Wed 31 Mar 2021 00:56:57 No.12905910 Report

Quoted By: >>12906384

>>12893521
Good resources on quantum computers?

Anonymous

Anonymous Wed 31 Mar 2021 03:15:42 No.12906384 Report

Quoted By: >>12907010

>>12905910
Nielsen and Chung (the “Mike’n’Ike” book)
Kitaev’s classical and quantum computation
And the link below
https://www.cs.umd.edu/~xwu/mini_lib.html

Anonymous

View Same Google iqdb SauceNAO 9E4FF534-1D65-4A8B-B0DC-12D0CF0A (...).png, 316KiB, 1200x738

Anonymous Wed 31 Mar 2021 03:40:23 No.12906450 Report

Quoted By:

>>12905588
I have found your worst nightmare

Anonymous

Anonymous Wed 31 Mar 2021 05:16:06 No.12906751 Report

Quoted By:

>>12900895
Nonsense

Anonymous

Anonymous Wed 31 Mar 2021 07:01:04 No.12907010 Report

Quoted By: >>12907030 >>12921091

>>12906384
Random question, I haven't really looked at books on quantum computing. But do they have a theoretical logical framework for it? Or do they rely just on quantum physics?

Anonymous

Anonymous Wed 31 Mar 2021 07:10:04 No.12907030 Report

Quoted By: >>12907108

>>12907010
>But do they have a theoretical logical framework for it? Or do they rely just on quantum physics?
What do you mean? If you're asking if there are computational models abstracted away from the physics, then yes. You will *eventually* want to know the physics, especially for a lot of applications of interesting theory, but you can get away at first thinking about quantum computing entirely in the language of vector spaces with complex probability amplitudes. For a decent amount of work, this is fine. But you will probably want to go deeper later on to tackle the big open problems (approximate ground state projection, quantum chem and qubitization, quantum chaos on graphs aka random matrix theory, entanglement and all of its exotic informatic meaning, etc etc..)
If you're interested in hardware, it should go without saying that you should look into physics and EE, particularly optics.

Anonymous

Anonymous Wed 31 Mar 2021 07:29:48 No.12907070 Report

Quoted By:

Measure theory from theoretical CS!
https://simons.berkeley.edu/talks/michael-mislove-2016-08-30

Anonymous

Anonymous Wed 31 Mar 2021 07:49:11 No.12907108 Report

Quoted By: >>12907307

>>12907030
That's what I mean yes, and I guess I need to learn probability theory it sounds like?

Anonymous

Anonymous Wed 31 Mar 2021 09:11:26 No.12907307 Report

Quoted By: >>12907419

>>12907108
Lmao yes, there are inherently probabilistic ideas associated to the quantum setting, many of which having to do with with fault tolerance / error correction and measurement. The former is one of the hottest fields of research on quantum computers, as it’s a theory topic directly tied to the effective implementation in hardware.
However, unless you’re super interested in quantum information theory, you don’t need the full weight of measure theory for a lot at first. Later there’s a lot of it though - the Haar measure ends up showing a lot lmao.

Anonymous

Anonymous Wed 31 Mar 2021 10:08:17 No.12907419 Report

Quoted By: >>12907964 >>12910315

>>12907307
I know I'm stupid. I'm just not the biggest fan of analysis, that's why I never took measure theory.

Anonymous

Anonymous Wed 31 Mar 2021 13:19:32 No.12907849 Report

Quoted By: >>12908316

Any good books about non relational databases?

Anonymous

Anonymous Wed 31 Mar 2021 14:11:15 No.12907964 Report

Quoted By:

>>12907419
you're not stupid. you can forego the measure theoretic probability for a good amount of things if it's not your forte.

Anonymous

Anonymous Wed 31 Mar 2021 14:35:51 No.12908036 Report

Quoted By: >>12908295

this thread and /mg/ are the best threads on /sci/

Anonymous

Anonymous Wed 31 Mar 2021 16:04:01 No.12908295 Report

Quoted By:

>>12908036
Glad to hear you enjoy these anon! If you want, you can suggest a potential topic for the next iteration

Anonymous

Anonymous Wed 31 Mar 2021 16:11:35 No.12908316 Report

Quoted By: >>12913322

>>12907849
We have a library down our way. All the books are on databases with no relations in them whatsoever. It is covid-closed these days, and actually there weren't many loaners even before.

Anonymous

Anonymous Wed 31 Mar 2021 16:43:51 No.12908434 Report

Quoted By: >>12908441

does sota discussion of deep learning belongs in here like transformers, self supervised learning and etc?

Anonymous

Anonymous Wed 31 Mar 2021 16:44:41 No.12908441 Report

Quoted By:

>>12908434
Go ahead anon. It’s fine

Anonymous

Anonymous Wed 31 Mar 2021 18:11:08 No.12908787 Report

Quoted By: >>12908819

anyone here into algorithmic game theory?
got an optional module on it next year I'm thinking of picking up but wanna know what i'm getting into first, so if anyone has any knowledge on it lemme know so i can ask some questions

Anonymous

Anonymous Wed 31 Mar 2021 18:17:18 No.12908819 Report

Quoted By: >>12921862

>>12908787
I don’t know much about it, but I think I’ve heard some stuff about it because of the unique games conjecture and its relevance to computing theory? That, and I’m pretty sure that that game theoretic microeconomics encountered a huge boom when algorithmic game theory entered the scene.

Anonymous

Anonymous Wed 31 Mar 2021 23:42:01 No.12910315 Report

Quoted By:

>>12907419
Folland is good for measure if you don’t like papa Rudin

Anonymous

View Same Google iqdb SauceNAO DAA431C0-6A84-4874-A7A9-2AB1859C (...).jpg, 172KiB, 887x1128

Anonymous Thu 01 Apr 2021 00:53:29 No.12910550 Report

Quoted By: >>12910607

Free survey book on the theory of computation and how math and CS have intersected for a while. Particularly, while we know CS uses math, it provides examples in the middle chapter “Interlude” and more of how CS has contributed to mathematics.
https://www.math.ias.edu/avi/book

It was written by one of the recent Abel Prize winners

Anonymous

Anonymous Thu 01 Apr 2021 01:12:07 No.12910607 Report

Quoted By: >>12911143

>>12910550
It's sad how so many CS programs have become just glorified programming schools. I really wish they'd separate the disciplines better, since now people are all majoring in CS when they just want to learn to program.

Anonymous

Anonymous Thu 01 Apr 2021 03:53:26 No.12911143 Report

Quoted By:

>>12910607
Eh, as it gets more mature, the super interesting graduate / Simon’s material will descend into undergraduate topics, and CS will be much more well defined as an undergrad curriculum. As the demand for engineers in the field rise, they’ll either be stratified to another specialized major made to simply write production code, or they will have to adapt to the CSE curriculum and get their practical experience through co ops and internships.

Basically, i think more schools will eventually follow the U Waterloo model of “give them a lot of tracks, don’t hold back on the material, and make sure they do lots of internships for practical experience”>>12910607

Anonymous

Anonymous Thu 01 Apr 2021 06:53:37 No.12911504 Report

Quoted By:

Book reccs from cstheory stackexchange
https://cstheory.stackexchange.com/questions/3253/what-books-should-everyone-read

Anonymous

Anonymous Thu 01 Apr 2021 09:05:17 No.12911784 Report

Quoted By:

The co-inventor of graph limits and graphons, analytic objects that show up quite a bit in statistics and math, and in property testing in CS
http://christianborgs.com/

Anonymous

Anonymous Thu 01 Apr 2021 13:24:54 No.12912361 Report

Quoted By: >>12912362

Can you guys help me with my 2nd sem undergrad assignment

Anonymous

Anonymous Thu 01 Apr 2021 13:26:50 No.12912362 Report

Quoted By: >>12912433

>>12912361
let's see it then

Anonymous

View Same Google iqdb SauceNAO Screenshot 2021-04-01 214856.png, 240KiB, 1092x624

Anonymous Thu 01 Apr 2021 13:52:43 No.12912433 Report

Quoted By: >>12912516 >>12912817 >>12916777 >>12917215

>>12912362
How am I drawing a blank after learning this stuff through 12 weeks. I'm just skimming through Kenneth H.Rosen's "Discrete Maths and its Applications" reference book now cuz it's the book used by the lecturer. How should I approach this?

Anonymous

Anonymous Thu 01 Apr 2021 14:21:55 No.12912516 Report

Quoted By: >>12912560

>>12912433
I am guessing the first one is a graph problem

Anonymous

Anonymous Thu 01 Apr 2021 14:37:16 No.12912560 Report

Quoted By: >>12912571

>>12912516
And adding to this astute observation, what makes it even more likely is that the second paragraph explicitly asks for a "graph model".

Anonymous

Anonymous Thu 01 Apr 2021 14:39:32 No.12912571 Report

Quoted By: >>12914200

>>12912560
these words suddenly kinda make sense now thanks anon. but im still retarded for the rest of the questions

Anonymous

Anonymous Thu 01 Apr 2021 15:56:22 No.12912817 Report

Quoted By: >>12916777

>>12912433
The second question is weird. First and foremost, they seem to overload the word “successful” here. That is, success seems to refer both to whether QA marks down the phone as being good or bad, and then success also means the confidence we have in QA’s affirmations / denials to be true. So the expected number of successes is 600k * .85 + 400k * .25 = 610k

When they say “construct an algorithm,” do they mean just to have some algorithm spit out the conditional probabilities given whether QA liked it or not?

Anonymous

Anonymous Thu 01 Apr 2021 19:41:09 No.12913322 Report

Quoted By:

>>12908316
??

Anonymous

Anonymous Fri 02 Apr 2021 00:21:30 No.12914200 Report

Quoted By:

>>12912571
>2nd semester
don't want to be rude, but even my highschool dropout brain can see that it's asking you for algorithm to compute and traverse hamiltonian path

Anonymous

Anonymous Fri 02 Apr 2021 00:35:06 No.12914231 Report

Quoted By: >>12914649 >>12915565 >>12916029

>when you're working on a simple machine learning project and spend ~3 hours trying to get your fucking logistic regression to work with your word vectors before you realise you only had to change the data structure containing them from a numpy array to a python list because scikit-learn changed since you last used it
yeah i hate it when this happens. i really do. i need to sleep

Anonymous

Anonymous Fri 02 Apr 2021 03:25:56 No.12914649 Report

Quoted By: >>12915565

>>12914231
practical ML usually takes a decent amount of fiddling with the libraries

Anonymous

Anonymous Fri 02 Apr 2021 12:06:35 No.12915565 Report

Quoted By: >>12916271

>>12914649
>>12914231
just make your own library lmao

Anonymous

Anonymous Fri 02 Apr 2021 15:04:23 No.12916029 Report

Quoted By: >>12916054

>>12914231
>using scikit-learn
use any autograd library (like pytorch or jax) or do it from scratch

Anonymous

Anonymous Fri 02 Apr 2021 15:10:56 No.12916054 Report

Quoted By: >>12917293

>>12916029
done it from scratch plenty of times (for my coursework at uni i wasn't allows to use libraries). at this point i understand the theory very well and just wanna get certain things done relatively quickly

Anonymous

Anonymous Fri 02 Apr 2021 16:19:50 No.12916271 Report

Quoted By: >>12917225

>>12915565
>reinvent the wheel, that'll learn em

Anonymous

Anonymous Fri 02 Apr 2021 18:26:48 No.12916777 Report

Quoted By:

>>12912817
>>12912433
any progress anon?

Anonymous

Anonymous Fri 02 Apr 2021 20:17:54 No.12917215 Report

Quoted By:

>>12912433
How is the first part here not impossible? By entering through C you leave only one valid path to enter B (or D) without backtracking.

Anonymous

Anonymous Fri 02 Apr 2021 20:19:55 No.12917225 Report

Quoted By:

>>12916271
The most infuriating part is how hard your sarcasm is ignored by the vast bulk of the industry.

Anonymous

Anonymous Fri 02 Apr 2021 20:34:43 No.12917293 Report

Quoted By: >>12917428

>>12916054
The whole point of a library is that it helps you get it done quickly - if that is not the case, then the existing libraries are quite poor. Given how many people are in the space and how many big names are contributing, I'd say that it's a reflection of how not-advanced the field really is.

Anonymous

Anonymous Fri 02 Apr 2021 21:05:55 No.12917428 Report

Quoted By:

>>12917293
>I'd say that it's a reflection of how not-advanced the field really is.
I think it's less that and more how it's focused on funneling for very specific applications. There's a lot of work into doing interesting application of ML, but there's simply a lack of manpower. You need people at least slightly more informed than the average codemonkey to do ML engineering and make as robust tools as what everyone else is using, or else they'll just fall back to scrubbing their data in order to work with the existing infrastructure.

Anonymous

Anonymous Sat 03 Apr 2021 03:35:00 No.12918859 Report

Quoted By:

Anyone here interested in chemistry and VQE (Variational Quantum Eigensolvers)?
https://qiskit.org/textbook/ch-applications/vqe-molecules.html

Anonymous

Anonymous Sat 03 Apr 2021 14:05:02 No.12920211 Report

Quoted By: >>12924528 >>12931769

Is the post quantum crypto guy here?

Anonymous

Anonymous Sat 03 Apr 2021 14:49:03 No.12920358 Report

Quoted By: >>12920450 >>12921601

yall hating codemonkey CS degrees but far worse is half assed math courses out of your specialty.

Anonymous

Anonymous Sat 03 Apr 2021 15:18:55 No.12920450 Report

Quoted By: >>12923554

>>12920358
lmao what? Just because "discrete math" as intro to proofs can be watered down at bad schools doesn't mean the math in CS is watered down.

Anonymous

Anonymous Sat 03 Apr 2021 18:00:33 No.12921091 Report

Quoted By:

>>12907010
you will eventually have to learn the physics

Anonymous

Anonymous Sat 03 Apr 2021 20:14:22 No.12921601 Report

Quoted By: >>12923570 >>12923701

>>12920358
I stopped hating code monkey once I realize that my web-dev friend has to update his knowledge on new frameworks every year to keep up. Web development has a major change in backend and frontend every 2 years at most.

Anonymous

Anonymous Sat 03 Apr 2021 21:30:47 No.12921862 Report

Quoted By: >>12922489 >>12923066

>>12908819
Unique games conjecture has nothing to do with game theory. It is a result for approximation theory of NP-hard problems.
UGC that basically says that it is impossible to approximate hard problems better than some bound. For example, if UGC is true, you cannot approximate vertex cover better than a factor of 2.

Anonymous

Anonymous Sun 04 Apr 2021 00:26:35 No.12922489 Report

Quoted By:

>>12921862
hmm, gotcha

Anonymous

Anonymous Sun 04 Apr 2021 03:32:16 No.12923066 Report

Quoted By: >>12923384

>>12921862
any resources to learn about UGC?

Anonymous

Anonymous Sun 04 Apr 2021 05:08:25 No.12923384 Report

Quoted By: >>12944933

>>12923066
This is a good writeup:
>>12923066" target="_blank">https://cs.nyu.edu/~khot/papers/UGCSurvey.pdf>>12923066
But I still recommend going back to read the original paper by Subhash Khot. You will probably need some background on PCP theorem. It is included in the writeup above, but it is explained better in some chapters in the Arora and Barak book.

Anonymous

Anonymous Sun 04 Apr 2021 06:08:14 No.12923554 Report

Quoted By:

>>12920450
You interpret it wrong. Implication is that i prefer codemonkey CS program over program filled with half assed math courses and shitty profs which is mostly to fullfil some faculty quota.

Anonymous

Anonymous Sun 04 Apr 2021 06:19:14 No.12923570 Report

Quoted By: >>12923701

>>12921601
Yeah webdev is changing constantly.
The problem is it's changing because of contrarians that keep changing all the time.
Yeah I get that this library does everything, but by the time you implement and learn it, I can code up the equivalent.
Obviously there are times DIY is retarded. But that's the difficulty of designing systems and allocating time.

Software development is a very different world to computer science.

Anonymous

Anonymous Sun 04 Apr 2021 07:35:50 No.12923701 Report

Quoted By: >>12925036 >>12930934

>>12921601
>>12923570
Don't dramatize, fluctuating softeng is a meme. DIY retardation and miriads of solutions to nonexistent problems means nothing because end user doesn't care about what's it's inside and some company will only hire you to work on their stack.
How different is it to use Spring or JQuery. Is there a way to work with databases that obliterates everything else? There are edge cases but they're not that difficult to comprehend.
The ones that meme webdev as being hard and major case of burnout is probably retarded pizza couriers who payed for one week bootcamp.

Anonymous

Anonymous Sun 04 Apr 2021 13:54:12 No.12924528 Report

Quoted By:

>>12920211
Still waiting for him

Anonymous

Anonymous Sun 04 Apr 2021 16:46:48 No.12925036 Report

Quoted By:

>>12923701
Keeping up with new technologies isn't hard unless you are lazy.
Softdev for junior and senior devs is only hard in that your customers might be retarded and they don't know what they wanted so you ended up with more work for some allotted time.
That's why I see people crunch a lot.
But if you are a team lead you will probably need some solid knowledge on everything and problem solving skills. That's why they are paid nice salary.

Anonymous

Anonymous Sun 04 Apr 2021 18:11:22 No.12925411 Report

Quoted By: >>12935043

>>12893521
I keep seeing "circuit complexity" show up in black hole physics and hep-th papers when searching on arXiv. Can any physicsfags elucidate what the physics community has been doing with this part of CS research, or if it means the same thing / analogous things at all?
All I know is that Leonard Susskind and Scott Aaronson apparently talk quite a lot

Anonymous

Anonymous Sun 04 Apr 2021 18:36:05 No.12925529 Report

Quoted By: >>12925544

anyone here somewhat well versed in the intersection of (mathematical) computer science and cognitive science? have one or two questions

Anonymous

Anonymous Sun 04 Apr 2021 18:39:52 No.12925544 Report

Quoted By:

>>12925529
not too well versed, but as far as I know, most cogsci and CS intersect at some parts of learning theory, optimization theory (talking about how the brain solves certain intractable tasks, of course computational neuroscience, etc etc..

Anonymous

Anonymous Sun 04 Apr 2021 23:09:07 No.12926790 Report

Quoted By:

Does this count as vision?
http://kurlin.org/projects/persistent-skeleton-dim2.pdf

Anonymous

Anonymous Mon 05 Apr 2021 05:36:20 No.12928079 Report

Quoted By: >>12928212 >>12929238 >>12929910 >>12929983

Since this thread has been in and out of the last page on the catalog, anyone have any suggestions for the topic of the next thread?

Anonymous

Anonymous Mon 05 Apr 2021 06:15:23 No.12928212 Report

Quoted By: >>12928284

>>12928079
I think people can just post their homework to keep it going.

Anonymous

Anonymous Mon 05 Apr 2021 06:36:08 No.12928284 Report

Quoted By:

>>12928212
lmao sure

Anonymous

Anonymous Mon 05 Apr 2021 11:46:42 No.12929238 Report

Quoted By: >>12929780

>>12928079
$z = f(\vec{x}; W, \vec{b}) = W_ix^i + \vec{b}$
$\hat{y} = max\{0, z\}$

Anonymous

Anonymous Mon 05 Apr 2021 14:54:03 No.12929780 Report

Quoted By: >>12929930

>>12929238
?

Anonymous

Anonymous Mon 05 Apr 2021 15:33:04 No.12929910 Report

Quoted By:

>>12928079
Don't worry about too much. Even /mg/ is slow as fuck recently (thread lasts for 5 days - a week).
It's probably because everyone is preparing for their finals.

Anonymous

Anonymous Mon 05 Apr 2021 15:36:52 No.12929930 Report

Quoted By: >>12930845 >>12930884

>>12929780
neural network nigga

Anonymous

Anonymous Mon 05 Apr 2021 15:50:13 No.12929983 Report

Quoted By: >>12930092 >>12930389

>>12928079
lol I always have questions regarding research and preparation for graduate school in TCS, but I don’t want to turn this into an advice thread. as far as a potential topic, my vote goes to one of
>Quantum Information and Computing
> Randomized and Probabilistic Algorithms

Anonymous

Anonymous Mon 05 Apr 2021 16:16:22 No.12930092 Report

Quoted By:

>>12929983
> I don’t want to turn this into an advice thread
Eh, I wouldn't mind giving advice. Mine might be limited though since I'm gonna be applying next sem

Anonymous

Anonymous Mon 05 Apr 2021 17:35:47 No.12930389 Report

Quoted By:

>>12929983
>but I don’t want to turn this into an advice thread.
It's okay as long as it's related to CS.

Anonymous

Anonymous Mon 05 Apr 2021 19:44:42 No.12930845 Report

Quoted By: >>12935110

>>12929930
what homework

Anonymous

Anonymous Mon 05 Apr 2021 19:53:42 No.12930884 Report

Quoted By: >>12935110

>>12929930
Braindead idiot and a nigger

Anonymous

Anonymous Mon 05 Apr 2021 20:02:29 No.12930934 Report

Quoted By:

>>12923701
I don't work for a company though.
If you're doing everything yourself it is not trivial.
As keeping up with technology is not just about "getting the newest". But also evaluating new options that all have balanced positives and negatives.

The bigger the company, the less responsive to change.
Keeping up within a company isn't hard. But keeping up for yourself to make for people is not easy.

Anonymous

Anonymous Mon 05 Apr 2021 21:01:02 No.12931313 Report

Quoted By: >>12931537

>Be a 3rd year CS PhD student
>No paper published
>Take a look at my friends and they are all publishing like crazy
I feel like shit.

Anonymous

Anonymous Mon 05 Apr 2021 21:43:59 No.12931537 Report

Quoted By: >>12931570 >>12932400

>>12931313
What subfield of CS?

Anonymous

View Same Google iqdb SauceNAO Untitled.png, 83KiB, 987x599

Anonymous Mon 05 Apr 2021 21:50:51 No.12931570 Report

Quoted By: >>12931595 >>12931684

>>12931537
algorithmic gay theory

Anonymous

Anonymous Mon 05 Apr 2021 21:55:12 No.12931595 Report

Quoted By:

>>12931570
Algayrithmic

Anonymous

Anonymous Mon 05 Apr 2021 22:15:04 No.12931684 Report

Quoted By: >>12931759

>>12931570
What books would you recommend for algorithmic game theory?

Anonymous

Anonymous Mon 05 Apr 2021 22:31:56 No.12931759 Report

Quoted By:

>>12931684
sorry i can't really help you there, my interests lie solely in algorithmic gay theory

Anonymous

Anonymous Mon 05 Apr 2021 22:34:15 No.12931769 Report

Quoted By: >>12932115

>>12920211
I'm not him but I am doing a project in post quantum cryptography. I'm looking at SIDH: https://en.m.wikipedia.org/wiki/Supersingular_isogeny_key_exchange

Anonymous

Anonymous Mon 05 Apr 2021 23:38:52 No.12932115 Report

Quoted By: >>12932635

>>12931769
Oooh nice, any PDFs/papers for reference?

Anonymous

Anonymous Tue 06 Apr 2021 00:45:56 No.12932400 Report

Quoted By: >>12932611

>>12931537
Combinatorial optimization.

Anonymous

Anonymous Tue 06 Apr 2021 01:44:16 No.12932611 Report

Quoted By:

>>12932400
Hmm, have you looked into Noga Alon’s work into certain relaxations of optimization problems? Admittedly I don’t know that much about combinatorial optimization aside from minimum spanning tree finding, almost perfect matching, and of course graph isomorphism. But I think you’d probably want to pick some other topics than graph iso if you need a publication soon

Anonymous

Anonymous Tue 06 Apr 2021 01:50:58 No.12932635 Report

Quoted By: >>12932799 >>12934891

>>12932115
Original paper: http://eprint.iacr.org/2011/506.pdf
Longer expository paper: https://arxiv.org/abs/1711.04062
Microsoft implementation: https://eprint.iacr.org/2016/413

Anonymous

Anonymous Tue 06 Apr 2021 02:39:28 No.12932799 Report

Quoted By: >>12933877

>>12932635
Thanks! What would the mathematics prereqs be for things in this field?

Anonymous

Anonymous Tue 06 Apr 2021 09:23:20 No.12933877 Report

Quoted By: >>12934533 >>12934891

>>12932799
The background for SIDH is quite specific and not really related to post-quantum stuff. You need a relatively large amount of algebraic geometry (divisors, Riemann-Roch, function fields, etc) and also more specific elliptic curve stuff. Otherwise some quantum stuff would probably be useful.

Anonymous

Anonymous Tue 06 Apr 2021 14:02:52 No.12934533 Report

Quoted By:

>>12933877
Got it. Thanks!

Anonymous

Anonymous Tue 06 Apr 2021 14:23:54 No.12934619 Report

Quoted By:

>>12893521
>This thread is dedicated to the mathematical aspects of computation.
Finally, a thread about real math.

Anonymous

Anonymous Tue 06 Apr 2021 14:36:33 No.12934654 Report

Quoted By: >>12934741 >>12934891

How can stuff like Differential Geometry and Algebraic (Topology or Geometry) be applied on AI (not just ML and shieeet)?

Anonymous

Anonymous Tue 06 Apr 2021 14:53:10 No.12934741 Report

Quoted By: >>12934945

>>12934654
same question but in CS overall, not just “AI”

Anonymous

Anonymous Tue 06 Apr 2021 15:34:42 No.12934891 Report

Quoted By: >>12934945 >>12935687 >>12935818 >>12937179

>>12934654
>How can stuff like Differential Geometry
We talked about this extensively in the last thread. Many of the preliminary resources are in the OP from Keenan Crane's courses / book on it
>>12841837
Also here are some notes from Albert Chern's course on it, which is also excellent (and his book is definitely a bit more in depth, with lots of classic and discrete differential geometry and a more than decent amount of physical examples / homework questions):
https://cseweb.ucsd.edu/~alchern/teaching/DDG.pdf
The course:
https://cseweb.ucsd.edu/~alchern/teaching/cse274_2020/
>Algebraic Topology
Lots of applications. One of the big ones is topological data analysis, which is a hodgepodge of a lot of topics in math.
Also look at Jeff Erickson's work
>Algebraic geometry
lots of applications / theory in CS that's been developed. On one hand, we saw a guy earlier on this thread talk about cryptography: SIDH and other quantum /post quantum schemes use quite a bit of alg geo though SIDH seems to use it the most >>12932635 >>12933877
And algebraic geometry is used heavily in geometric complexity theory and algebraic complexity in general. Here are some good resources.
A course:
http://pcwww.liv.ac.uk/~iken/teaching_sb/summer17/introtogct/index.html
Its lecture notes:
http://pcwww.liv.ac.uk/~iken/teaching_sb/summer17/introtogct/gct.pdf
Landsberg's short form intro for people already in algebraic geometry:
https://arxiv.org/abs/1305.7387
Landsberg's full text on it (very extensive):
https://www.math.tamu.edu/~jml/simonsclass.pdf

Anonymous

Anonymous Tue 06 Apr 2021 15:48:17 No.12934945 Report

Quoted By:

>>12934891
meant to answer >>12934741

Anonymous

Anonymous Tue 06 Apr 2021 16:05:51 No.12935043 Report

Quoted By:

>>12925411
maybe I should ask this in a physicsfag thread

Anonymous

Anonymous Tue 06 Apr 2021 16:19:38 No.12935110 Report

Quoted By:

>>12930845
>>12930884
>anon asks recommendation for the next thread topic
>recommends neural network
>called braindead nigger idiot with a homework
??? what part of neural network that isnt mathematical computer science

Anonymous

Anonymous Tue 06 Apr 2021 18:37:02 No.12935629 Report

Quoted By: >>12935678 >>12935697 >>12935732 >>12948453

https://boards.420chan.org/prog/thread/39023/functionalcs2-electric-boogaloo

What do you anons think of this guy? He shills Wildberger constantly, but he does say a lot of interesting stuff.

Anonymous

Anonymous Tue 06 Apr 2021 18:49:04 No.12935678 Report

Quoted By: >>12935697 >>12935818 >>12936441 >>12936995

>>12935629
He’s a pseud who does the “online learning community” thing where he talks about doing hard content in an incredibly ad-hoc, “we’ll learn as we go” approach. You should be incredibly skeptical of anybody who claims they can take a topic typically for grads and learn the essentials as a total beginner. The only reason people like to do this with ML is because normies and brainlets see $$$ and think it’s just as braindead as frontend software dev. It’s fine to learn AI from scratch. It’s not fine to do it before you’ve ever written a line of code or diagonalized a matrix.

You do need to wing it to some extent, but there’s a reason why the PhD courses are separated from the undergrad ones, and why PhD ML courses say you should “at the least” be okay with calculus, linear, analysis, etc and have a good amount of mathematical maturity. The real preq structure would be extensive, but you’re expected to wing it and learn the relevant functional analysis, measure theory, etc. on demand. But this also prompts you to learn it *properly.* The 420chan guy strikes me as the type of pseud who reads over everything at a cursory glance and invites a community of people to watch him try and convince himself (and everyone else) that he’s a smart cookie.

Anonymous

Anonymous Tue 06 Apr 2021 18:52:44 No.12935687 Report

Quoted By:

>>12934891
wow these recs are great.

Anonymous

Anonymous Tue 06 Apr 2021 18:55:50 No.12935697 Report

Quoted By: >>12935818

>>12935629
>>12935678
Also be very wary of anybody who unironcally shills for wildberger. There isn’t anything wrong taking an issue on irresponsible use of infinity in mathematics, but the takeaway that most people get from wildberger is to throw the baby out of with the bathwater. They’ll parrot views about computability despite having no experience with classic computability and recursion theory, which is more than comfortable with infinite objects. If we’re talking about decidable by electronic computers, even then there are results like how Escardo showed certain infinite sets admit exhaustive finite time search. Dealing with infinite objects is the job of any mathematician, physicist, computer scientist, etc.. but the key point is putting a canonical ontology to it via context. We can make sense of analysis by thinking about arbitrary precision and approximation. We can make sense of smoothness vs. turning angle, internal angular sum, Steiner formula, etc.. and other objects.

Anonymous

Anonymous Tue 06 Apr 2021 19:04:45 No.12935732 Report

Quoted By: >>12935818 >>12936441 >>12936732 >>12936995

>>12935629
>If you don't know what you want to do, and want direction, I suggest you email RMS
Is it really in question if this guy is a poser pseud who has very little idea what he’s talking about? RMS hasn’t had anything to do with math or mathematical CS since his Harvard undergrad, much less even free software, which was a big systems CS / software engineering project. He knows even less about AI. RMS hasn’t been relevant in CS for decades, even if free software has.
This dude is a /g/ tier faggot.
>this time doing algorithm optimization meaning I project collected data using Wildberger rational math, which then brings you into the realm of TAOCP and Kozen's Algorithm book where it's very easy to optimize,
Correction: this guy is a /g/ tier faggot who spouts buzzwords from schizo threads on /sci/

Anonymous

Anonymous Tue 06 Apr 2021 19:30:25 No.12935818 Report

Quoted By: >>12937197

>>12934891
Those are the stuff that I am interested in looking at, thanks!
>>12935697
>>12935732
>>12935678
I have the nagging feeling that he was a pseud the more that I gained the background knowledge in math and CS. I kinda followed his curriculum back then, until he started jumping around like an ADHD-addled retard on different resources.
Though I wish I realized earlier kek. It's much better to just go to another universities' curriculum or just dive right into a textbook rather than following other's progress.

What's the best way anyways to establish some kind of correspondence with other researchers in TCS? Most research in CS here are kinda meme-y (trivial applications or the famous buzzword fields).

Anonymous

Anonymous Tue 06 Apr 2021 21:46:54 No.12936415 Report

Quoted By:

>>12900657
any anons in here know of any interesting applications of modal logic in computer science? I need to come up with a proposal for a term paper for my modal logic course. we unfortunately didn't discuss any applications in the course, but I know they are out there.

Anonymous

Anonymous Tue 06 Apr 2021 21:50:42 No.12936441 Report

Quoted By: >>12936592 >>12936732

>>12935678
>He’s a pseud who does the “online learning community” thing where he talks about doing hard content in an incredibly ad-hoc, “we’ll learn as we go” approach. You should be incredibly skeptical of anybody who claims they can take a topic typically for grads and learn the essentials as a total beginner.
Where does he do that? The resources he proposes in "from scratch" series is the most basic shit you can think of.
>It’s not fine to do it before you’ve ever written a line of code or diagonalized a matrix.
Okay I kind of agree but again could you show me where he does that?
>>12935732
Do you sit in RMS's head you know what he might or might not know?

Anonymous

Anonymous Tue 06 Apr 2021 22:18:53 No.12936592 Report

Quoted By: >>12936995

>>12936441
>Where does he do that?
His intro post is about jumping from Learn AI from Scratch, which itself is ok but not the best intro to programming and intro to CS course ever, but w/e, enthusiasm at that level is more important than covering all your bases.
It's the point where he says that "we're" (which I assume is his thread or community) gonna follow the lectures and coursework of 2 PhD level ML courses (should be around the third paragraph) that I raise my eyebrow. He says that it doesn't matter if you're a beginner, that you can do it. He repeats that a lot.
>Do you sit in RMS's head you know what he might or might not know?
Anon, this is an argument done in bad faith and you know it. By this, we can't know ANYTHING that anybody might know because they could be withholding it, but I can make a judgement call based on
>RMS's involvement (or really lack thereof) in any technical capacity with free software
>RMS's blog
>RMS's work at large, which is basically just being a figurehead when he's not being a lazy douche
as well as RMS's lack of ANY demonstrated background or work in ML/AI. He did well in the Harvard math 50 course as an undergrad, so it's safe to say he was great at mathematics at some point, but again, given that all of his work has skewed into software engineering and laying out patents, I can say with confidence that I don't think RMS would be any sort of practical help in learning ML or getting inspired ideas.

As far as RMS goes in general, I don't think there's anything you'd get from emailing him anyway. He's mostly revered by pseud hipsters online who think shallow skepticism towards tech conglomerates (which is justified) equips them with knowledge and novelty of thought to actually have anything meaningful to say about software and even CS. Basically, they have massive dunning kruger. Anyone who knows what they're talking has a bullshit meter that can detect washed up has beens who no longer has anything to of value to say.

Anonymous

Anonymous Tue 06 Apr 2021 22:57:46 No.12936732 Report

Quoted By:

>>12935732
based /g/ is a fucking shithole and rms is an old fart that never did anything impressive outside software engineering, but that said,

I want a /coding/ general that isn't shit

>>12936441
Commit seppuku /g/tard

Anonymous

View Same Google iqdb SauceNAO Screenshot 2021-04-07 005406.png, 24KiB, 1108x211

Anonymous Tue 06 Apr 2021 23:54:33 No.12936988 Report

Quoted By: >>12940635

anyone know much about adversarial machine learning? specifically with regards to debiasing models.

in other words, how the fuck do i train an objective that looks like this?

Anonymous

View Same Google iqdb SauceNAO books .jpg, 481KiB, 1329x898

Anonymous Tue 06 Apr 2021 23:54:47 No.12936991 Report

Quoted By: >>12937074 >>12942952 >>12943378

Which is better?

Anonymous

Anonymous Tue 06 Apr 2021 23:57:04 No.12936995 Report

Quoted By: >>12938373

>>12936592
>>12935678
>The 420chan guy strikes me as the type of pseud who reads over everything at a cursory glance and invites a community of people to watch him try and convince himself (and everyone else) that he’s a smart cookie.
Weirdly enough, You seem to be doing the exact thing you accuse him of. You just looked at the 1st post, and proceed to make a sweeping generalisation about the person.

>>12935732
His advice about emailing rms, wasn't to learn about AI from him. It was to get a general advice about the tech world. Kind of like how you'd ask bill gates for general direction regarding the tech world, with the only difference that rms bothers responding back to plebs like you and me, and gates doesn't.

Anonymous

Anonymous Wed 07 Apr 2021 00:21:33 No.12937074 Report

Quoted By:

>>12936991
they're a full year sequence written by the exact same authors. The blue book is a somewhat easy but good intro to analyzing algorithms. The purple book is way harder and requires familiarity and comfort with complex analysis to get the most out of it, though the first half is much more combinatorial. It's not an easy book though, and there aren't any exercises, so you'll have to find your own or other exercises online.

Anonymous

Anonymous Wed 07 Apr 2021 00:51:47 No.12937179 Report

Quoted By: >>12938529

>>12934891
Do you know any interesting applications of topological data analysis? I took a course in it and it seemed a lot of the prominent papers were just using it as feature extraction for their NN. There weren't really any problems it seemed uniquely good at solving.

Anonymous

Anonymous Wed 07 Apr 2021 00:57:39 No.12937197 Report

Quoted By: >>12938373

>>12935818
>I have the nagging feeling that he was a pseud the more that I gained the background knowledge in math and CS.
In what way?

Anonymous

Anonymous Wed 07 Apr 2021 01:27:43 No.12937286 Report

Quoted By: >>12938525

Anyone know how you would apply Poisson distribution to financial markets?
I heard some guy saying he used a mean regression strat with Poisson distribution to make bank but it's only used for discrete quantities, so I'm wondering what he was measuring with the distribution.
Maybe the number of green candles in a given period?

Anonymous

Anonymous Wed 07 Apr 2021 05:48:45 No.12938159 Report

Quoted By:

Heckin bumperino

Anonymous

Anonymous Wed 07 Apr 2021 07:02:42 No.12938373 Report

Quoted By: >>12938504

>>12936995
>an RMS fanboy here to play defense force
I am judging this person's call to action post, yes, because his goals are more than ambitious. They reek of project doomed to fail and self aggrandizing pseudointellectual bullshit for the aforementioned reasons.
>m. It was to get a general advice about the tech world. Kind of like how you'd ask bill gates for general direction regarding the tech world,
RMS hasn't had any takes about the tech world that haven't been room temperature for the past 20 years.
> that rms bothers responding back to plebs like you and me, and gates doesn't.
See above, and RMS has literally nothing else going on in his life. This isn't a defense of Gates, but that I don't think either of them would have anything insightful to say to some rando on a forum.
>>12937197
If you know anything about math past intro to proofs and anything about CS past basic complexity and intro AI, you can tell he's selling the idea of getting aggressively into AI through effort but really just surrounding this mission with a bunch of clearly unachievable goals, No, you cannot just jump into a PhD ML course with barely a background, much less no programming experience whatsoever. Learn AI from scratch is okay but not enough to "just fucking do it [academic ML at the PhD level] because you can." Suggesting otherwise is just stroking your dick - if you were serious, you'd start with a deep understanding of the fundamentals, even if abbreviated, and then jump in.

Anonymous

Anonymous Wed 07 Apr 2021 07:58:58 No.12938504 Report

Quoted By: >>12938738

>>12938373
>No, you cannot just jump into a PhD ML course with barely a background, much less no programming experience whatsoever.
Okay that sound reasonable, but could you show a quote of him saying that people should do that?

Anonymous

Anonymous Wed 07 Apr 2021 08:12:06 No.12938525 Report

Quoted By:

>>12937286
Technically all digital data is discrete.

Anonymous

Anonymous Wed 07 Apr 2021 08:13:08 No.12938529 Report

Quoted By: >>12940233

>>12937179
We use TDA in network distributions, image reconstruction, clustering etc.

Anonymous

Anonymous Wed 07 Apr 2021 10:08:30 No.12938738 Report

Quoted By: >>12939603

>>12938504
>https://boards.420chan.org/prog/thread/39023/functionalcs2-electric-boogaloo
>here's the deal: we are doing CMU's ML 10-601 and 10-701 phd courses in a week or so, from scratch, and it doesn't matter because you anon can do it too. Fuck it we'll just figure it out from scratch.
Bonus pseud ramblings:
> had to prove myself to get into the lab I'm currently larping in as a genetic researcher only because I had old letters of recommendations

Anonymous

Anonymous Wed 07 Apr 2021 15:18:44 No.12939603 Report

Quoted By: >>12955242

>>12938738
He might be a pseud but you anon need to either take your meds or sigh up yourself to some reading comprehension course

Anonymous

Anonymous Wed 07 Apr 2021 16:44:17 No.12939991 Report

Quoted By: >>12940640

Anyone here know anything about dimensionality reduction? Seems like that field is heavily applied functional analysis

Anonymous

Anonymous Wed 07 Apr 2021 17:31:51 No.12940233 Report

Quoted By:

>>12938529
Do you have any good books or papers to get into it? It seems like you need to be generally good at a lot of algebra and probability theory. I only know about barcodes and persistent homology though

Anonymous

Anonymous Wed 07 Apr 2021 19:03:17 No.12940635 Report

Quoted By:

>>12936988
Can you give me a context of the objective function?
Paper name?

Anonymous

Anonymous Wed 07 Apr 2021 19:04:21 No.12940640 Report

Quoted By: >>12940790

>>12939991
You mean things like PCA / features embedding?
I don't know anything about functional analysis but dimensionality reduction methods are usually used in stat / ML.

Anonymous

Anonymous Wed 07 Apr 2021 19:25:47 No.12940732 Report

Quoted By: >>12940756 >>12940897 >>12941192

I'm a math MSc but I've never studied theoretical CS before. I don't even know what comprises the area of theoretical CS. How should I start?

Anonymous

Anonymous Wed 07 Apr 2021 19:29:49 No.12940756 Report

Quoted By: >>12940784

>>12940732
Do you have some knowledge on algorithm and algorithm complexity at least?
If you do, start with the Arora Barak book.
Otherwise, go read CLRS to have some idea.

Anonymous

Anonymous Wed 07 Apr 2021 19:33:37 No.12940784 Report

Quoted By:

>>12940756
Yeah I do, thanks for the suggestion, will read it right away.

Anonymous

Anonymous Wed 07 Apr 2021 19:34:47 No.12940790 Report

Quoted By: >>12940824

>>12940640
Gotcha. I was reading through
https://people.eecs.berkeley.edu/~minilek/publications/papers/terminal_jl.pdf
and wanted to know some application areas or common places of intersection.

Anonymous

Anonymous Wed 07 Apr 2021 19:40:07 No.12940824 Report

Quoted By: >>12942540

>>12940790
That looks like a nice result and remind me of core set, could lead to interesting theory results in dimensionality reduction in ML but I'm not entirely sure.

Anonymous

Anonymous Wed 07 Apr 2021 19:52:18 No.12940897 Report

Quoted By: >>12940962

>>12940732
>Complexity
Arora-Barak
>Algo
If you have no experience in their *design*: CLRS is good
If you have no experience in their *analysis*:
introduction to Analysis of Algorithms by Flajolet and Sedgewick is good. The material is familiar, but it should essentially be how to use calc 2 and very basic analysis in clever ways.
If you have some experience in both and have some background in complex analysis:
Analytic Combinatorics is good

The theory is still developing and mostly contained in papers, but these are good references that'll let you touch a lot of the other topics.

Anonymous

Anonymous Wed 07 Apr 2021 20:04:40 No.12940962 Report

Quoted By: >>12941859

>>12940897
Great list, thank you anon.

Anonymous

Anonymous Wed 07 Apr 2021 20:46:17 No.12941192 Report

Quoted By:

>>12940732
connecting traditional math to TCS can literally be free money lmao

Anonymous

View Same Google iqdb SauceNAO B398D9E2-7E87-4893-A1D3-EBC6A3A8 (...).jpg, 499KiB, 1125x1256

Anonymous Wed 07 Apr 2021 22:38:39 No.12941859 Report

Quoted By: >>12941870

>>12940962
Also two common themes:
The first half of the books are all easy. They test core concepts that can be tricky to internalize but “behave” nicely. Not much math is needed except carefully reading definitions and occasionally being clever. This gives the impression that topics in CS are generally “not deep” and don’t take much math to study.

The last half / third of these books gets dramatically more difficult. Aside from analytic combinatorics, they still assume no specific math background, but the material much more closely resembles “traditional math.” Results are still developed within the text, but they assume the reader will be comfortable jumping in. Pic related isn’t very hard to prove, particularly if you have any math background, but it starts reaching back to familiar languages like Fourier analysis.

You can expect more math to come your way as you get closer and closer to the cutting edge. This means CS doesn’t have good “PR” to undergrads or even adjacent grad students that it’s a mathematical subject, but if you seek it, you will find.

Anonymous

Anonymous Wed 07 Apr 2021 22:40:16 No.12941870 Report

Quoted By:

>>12941859
>Aside from analytic combinatorics, they still assume no specific math background
Meant to say, aside from assuming some knowledge of complex analysis in analytic combinatorics, barely any math background aside from intro to proofs, linear algebra, and mathematical maturity is expected.

Anonymous

Anonymous Thu 08 Apr 2021 01:06:40 No.12942540 Report

Quoted By:

>>12940824
I think stuff like UMAP is used in ML no?

Anonymous

Anonymous Thu 08 Apr 2021 02:23:53 No.12942952 Report

Quoted By: >>12942993

>>12936991
Neither. CLRS is the only acceptable reference book.

Anonymous

Anonymous Thu 08 Apr 2021 02:34:24 No.12942993 Report

Quoted By:

>>12942952
CLRS is okay but the books you responded to are about analyzing algorithms, not designing them. The math in them is a lot more involved than in CLRS because they're made to help you understand more complicated arguments used in algorithm design and complexity.

Anonymous

Anonymous Thu 08 Apr 2021 04:49:11 No.12943378 Report

Quoted By:

>>12936991
They’re part 1 and part 2 of a unified text on the analysis of algos. Basically it’s applied combinatorics and complex analysis

Anonymous

Anonymous Thu 08 Apr 2021 05:06:18 No.12943426 Report

Quoted By: >>12943438 >>12943649 >>12943671 >>12948120

>>12893521
does complex analysis show up much in TCS?

Anonymous

Anonymous Thu 08 Apr 2021 05:10:19 No.12943438 Report

Quoted By: >>12943649 >>12943671

>>12943426
I don't think so. Signal processing I guess, but you pretty much only need the definition of complex numbers to understand the Fourier series. Anyway, most would consider that applied CS.

Anonymous

View Same Google iqdb SauceNAO 2EFD6B39-A93E-4122-8EE2-7DC7836D (...).jpg, 38KiB, 350x499

Anonymous Thu 08 Apr 2021 06:20:58 No.12943649 Report

Quoted By: >>12943671

>>12943426
>>12943438
Sort of but not everywhere. It’s been mentioned a few times in this thread that Analytic Combinatorics uses complex analysis, particular to make sense of the “transfer theorems.” Basically, certain combinatorial objects correspond to certain generating functions. Generating functions can be analyzed via complex analysis. The machinery of complex analysis proves a lot about these generating functions, particularly how to “compose” them together. This gives us a very “clean” way of composing different “classes” of objects to count. You can reduce a lot of very complicated counting questions by specifying some underlying structure via the counting classes (these are called “symbolic methods of enumeration”). Then the analytic transfer theorems tell you immediately how to count it.
To understand and use these results for hard questions, you need to know complex analysis. But not everyone needs to use these results.
Here are some other example of it showing up in theory:
https://cstheory.stackexchange.com/questions/15057/complex-analysis-in-theoretical-computer-science

As for signals, while they are more “applied,” Fourier analysis shows up quite a lot in theory CS as a way to understand Boolean functions. Here is a good survey on the celebrated “hypercontractivity” results and other things like KKL:
https://www.cs.tau.ac.il/~amnon/Classes/2016-PRG/Analysis-Of-Boolean-Functions.pdf

Anonymous

View Same Google iqdb SauceNAO EE9D310E-491E-4FFF-973A-6DD0456B (...).jpg, 582KiB, 1125x1464

Anonymous Thu 08 Apr 2021 06:30:45 No.12943671 Report

Quoted By:

>>12943426
>>12943438
>>12943649
Also pic related is relevant. Complex analysis isn’t needed until it is. Given it’s use in understanding key objects in math (and its historical use to characterizing the discrete, as in analytic number theory), I wouldn’t be surprised if it ends up showing more and more in TCS. One this is for sure though: while you can “get away” with not being good at analysis in a lot of areas in TCS, you’re seriously holding yourself back from a lot of the most interesting results if you aren’t goof at analysis. Highly recommend going through chapters 1-6 of baby Rudin at the very least to hone those skills with your inequalities and writing our definitions. Analytic arguments and thinking shows up more than you’d think.

Anonymous

Anonymous Thu 08 Apr 2021 14:05:15 No.12944933 Report

Quoted By: >>12945271

>>12923384
Any other recent work on it? I saw that people were able to prove the propeller conjecture in the k = 4 case about 7 years ago

Anonymous

Anonymous Thu 08 Apr 2021 15:35:57 No.12945271 Report

Quoted By:

>>12944933
I just read it to know the basic because it's pretty interesting but I haven't kept up with the latest work.

Anonymous

Anonymous Thu 08 Apr 2021 18:53:03 No.12946107 Report

Quoted By: >>12951086

We got randomized algo and quantum computing on the suggestions list right now. Any other topics anybody would be interested in? Here are a few possible ones:

Recursion theory / Computability
Robotics
Programming Language Theory and Compilers
Machine Learning
Bioinformatics and Bioengineering

Anonymous

Anonymous Thu 08 Apr 2021 20:20:03 No.12946532 Report

Quoted By: >>12947004 >>12947016

>>12893521
>CS
Not science or math.
>>>/g/

Anonymous

Anonymous Thu 08 Apr 2021 20:23:38 No.12946547 Report

Quoted By: >>12947011

What would be a good discrete math book to read before going into CLRS?

Anonymous

Anonymous Thu 08 Apr 2021 22:00:35 No.12947004 Report

Quoted By:

>>12946532
>waaaaah CS doesn’t have math
Why do you reee about codemokeying on threads that have nothing to do with codemonkeying?

Anonymous

Anonymous Thu 08 Apr 2021 22:02:17 No.12947011 Report

Quoted By: >>12947042

>>12946547
Discrete math is huge. I’d say it’s probably fine to just read intro to proofs. CLRS doesn’t have very strict mathematical background at all.
You should read analysis of algorithms by sedgewick and flajolet if you want more involved math in algo, and analytic combinatorics if you want to see the real “face” of algorithm analysis.
But don’t rush yourself

Anonymous

Anonymous Thu 08 Apr 2021 22:03:18 No.12947016 Report

Quoted By:

>>12946532
/g/ has nothing to do with CS

Anonymous

Anonymous Thu 08 Apr 2021 22:09:35 No.12947042 Report

Quoted By:

>>12947011
>CLRS doesn’t have very strict mathematical background
I see, I saw it being a rec to that math MSc, and thought it was heavier.
Thank you, gonna at least check those books.

Anonymous

Anonymous Fri 09 Apr 2021 02:29:56 No.12948120 Report

Quoted By:

>>12943426
TCS is a wild card. New math keeps showing up as time goes on and the field matures

Anonymous

Anonymous Fri 09 Apr 2021 03:06:40 No.12948207 Report

Quoted By: >>12949762

>>12898127
Nikolaj? I've seen people posting your stuff on /mg/, are you still on /sci/?

Anonymous

Anonymous Fri 09 Apr 2021 04:33:50 No.12948453 Report

Quoted By: >>12949174

>>12935629
Not to shit on this anon. He does introduce you to very good resources. But doing them the way he asks you to do it is very difficult if you're already in university. I've had more luck with his resources in my own way rather than using his approach.
The other thing is that he sometimes boasts that getting into firms like Jane Street is easy if you've done PAPL. I admittedly do not have much experience with HFT but that statement definitely sounds fake.
But kudos to that anon. He has clearly put effort into the curriculum and I got to know about many good resources because of that.

Anonymous

Anonymous Fri 09 Apr 2021 08:42:04 No.12949174 Report

Quoted By:

>>12948453
nah, shit on all pseuds. I don't believe him for a second

Anonymous

Anonymous Fri 09 Apr 2021 08:56:27 No.12949221 Report

Quoted By: >>12949227 >>12951119

I am currently trying to find out people who are working on creating a rigorous mathematical framework to formalize Deep Learning.

What I mean is like, the hidden layers are clearly some kind of transformation in some kind of space. What is this space and what are these transformations and how does backpropagation approximate them.

Does anyone know of anyone working on this stuff?

Anonymous

Anonymous Fri 09 Apr 2021 08:59:26 No.12949227 Report

Quoted By:

>>12949221
The space of uhhhh substructures in matter itself

Anonymous

Anonymous Fri 09 Apr 2021 09:01:10 No.12949237 Report

Quoted By:

>>12893521
All of it is incorrect. Its government psyops to destroy you. Toggle.

Anonymous

View Same Google iqdb SauceNAO 1617968545818.png, 97KiB, 1440x219

Anonymous Fri 09 Apr 2021 11:42:54 No.12949709 Report

Quoted By: >>12953161

Anonymous

Anonymous Fri 09 Apr 2021 12:03:19 No.12949762 Report

Quoted By:

>>12948207
That wasn't me. I don't think I've used cuda. I check /mg/ occationally

Anonymous

Anonymous Fri 09 Apr 2021 18:13:50 No.12951086 Report

Quoted By:

>>12946107
I can add more topics to this if people want

Anonymous

Anonymous Fri 09 Apr 2021 18:16:48 No.12951096 Report

Quoted By: >>12951136

>>12893521
>math
Come on now, nigger, if I was any good at math I wouldn't have picked computer science.

Anonymous

Anonymous Fri 09 Apr 2021 18:18:00 No.12951105 Report

Quoted By: >>12951140

>>12893922
>car designers
are literally engineers lmao

A better analogy would be physicists and car designers.

Anonymous

Anonymous Fri 09 Apr 2021 18:21:01 No.12951119 Report

Quoted By:

>>12949221
What exactly are you looking for? What makes you think you will ever find a satisfactory low-dimensional representation of whatever the fuck it is neural net is doing, when the reason we are using neural nets to begin with is that these problems are intractable when you try to reduce their dimensionality?

Anonymous

Anonymous Fri 09 Apr 2021 18:23:30 No.12951136 Report

Quoted By: >>12951963

>>12951096
This is the problem lmao. Computer science is inherently full of math, but they abstract it away at a lot of schools because math scary. But if you don’t like math, CS is the wrong thing to break into. Unless your idea of CS is entry level software

Anonymous

Anonymous Fri 09 Apr 2021 18:24:33 No.12951140 Report

Quoted By: >>12951963

>>12951105
There are definitely CS BSc’s who end up doing engineering work, as has been mentioned a few times in the thread. I think the problem is that anyone who touches software is suddenly thrust into the title of “engineer” which devalues the name

Anonymous

Anonymous Fri 09 Apr 2021 21:53:48 No.12951963 Report

Quoted By: >>12952509 >>12952515

>>12951136
nigger I'm in grad school
>>12951140
my point is that your analogy is terrible since engineers are the people who design cars, not the people who put them together on an assembly line or fix them at a shop

Anonymous

Anonymous Fri 09 Apr 2021 22:21:39 No.12952111 Report

Quoted By: >>12952527

Forgive me if this is a retarded question (I'm new to TCS).
If my understanding correct that Turing machines that are deciders are equivalent to computable functions, and Turing machines that aren't deciders are equivalent to partially computable functions?
My main query is whether equivalent is the correct word to use here.

Anonymous

Anonymous Sat 10 Apr 2021 00:10:08 No.12952509 Report

Quoted By: >>12955268

>>12951963
>his grad program in CS doesn't use math
you sound like either a systems brainlet or a masters student, not a PhD one lmao

Anonymous

Anonymous Sat 10 Apr 2021 00:12:54 No.12952515 Report

Quoted By: >>12955268

>>12951963
>i'm in grad school
oh you're the type of brainlets who try and remove CS away from math, got it lmao.

Anonymous

Anonymous Sat 10 Apr 2021 00:16:13 No.12952527 Report

Quoted By:

>>12952111
Turing machine, at its core definition, reads some input and outputs 0 or 1 (or halt halfway through). Say, you have a set of strings of specific syntax (i.e. contains only number 1), we say a TM decides that set of strings if it outputs 1 whenever it sees a string belongs to that set, and 0 (or halt) for anything else.
If a TM doesn't decide a set of strings, it can output 0, 1, or halt when it sees element from that set.

Anonymous

Anonymous Sat 10 Apr 2021 03:35:42 No.12953161 Report

Quoted By:

>>12949709
>if I can't solve it, it's probably unsolvable
based

Anonymous

Anonymous Sat 10 Apr 2021 06:55:22 No.12953716 Report

Quoted By: >>12953741

Remember your roots, and remember that we can actually study randomness, statistics, etc., with computers is a miracle of mathematics and computation.
Shot noise and electrical processes are cool, but generating distributions algorithmically is really fucking cool.
http://www.nrbook.com/devroye/

Anonymous

Anonymous Sat 10 Apr 2021 07:01:35 No.12953741 Report

Quoted By: >>12954849

>>12953716
Doesnt the only source of randomness in the computer come from quantum variances in really tiny electronics? Either that or the internet which is external inputs.

Anonymous

Anonymous Sat 10 Apr 2021 12:47:46 No.12954849 Report

Quoted By: >>12954850

>>12953741
Think pseudorandomness. Aside from cryptographic applications, you don’t really need true randomness a lot of the time.

Anonymous

Anonymous Sat 10 Apr 2021 12:49:15 No.12954850 Report

Quoted By: >>12956080

>>12954849
But wheres tat cone from

Anonymous

Anonymous Sat 10 Apr 2021 15:02:09 No.12955242 Report

Quoted By:

>>12939603
You're clearly the guy. why are you pretending not to be him?

Anonymous

Anonymous Sat 10 Apr 2021 15:09:36 No.12955268 Report

Quoted By:

>>12952509
>>12952515
It does have math, I just suck at it.

Anonymous

Anonymous Sat 10 Apr 2021 16:28:32 No.12955544 Report

Quoted By: >>12956086

Opinion on this program?
https://catalog.njit.edu/undergraduate/science-liberal-arts/mathematical-sciences/computational-mathematics/

Anonymous

Anonymous Sat 10 Apr 2021 18:19:12 No.12956080 Report

Quoted By:

>>12954850
Psuedorandomness is done algorithmically. You’re thinking of choosing a seed, which is an expensive physical operation to do correctly. They usually use something like thermal noise, with the assumption that it’s true random because it follows various distributions. You use a random seed (or even something foolish like the system time) as input to a pseudorandom number generator, which can give you a distribution of different values.
Pseudorandomness is clever because it’s a very specific middle ground between determinism and being “random enough.” The idea is that there’d be no way to actually verify in polynomial fine if the numbers aren’t random by the way they’re constructed. Done correctly, pseudorandomness is super powerful

Anonymous

Anonymous Sat 10 Apr 2021 18:20:32 No.12956086 Report

Quoted By: >>12956265

>>12955544
Not bad, seems to be focus on numerical methods for the sciences. Pick your electives wisely based on your interests

Anonymous

Anonymous Sat 10 Apr 2021 18:22:59 No.12956100 Report

Quoted By: >>12956130 >>12956334

>>12955687
made this thread but maybe I can get answered here

Anonymous

Anonymous Sat 10 Apr 2021 18:29:08 No.12956130 Report

Quoted By: >>12956164

>>12956100
Graphics is an open and active field of research. You’re talking about a very specific subset of graphics research though aren’t you? Though it’s hard to distinguish since afaik (im not really part of the graphics community) rendering is still active if you look at popular SIGGRRAPH papers.

Remember, not easy for middle ground companies to implement != dead field

Anonymous

Anonymous Sat 10 Apr 2021 18:36:09 No.12956164 Report

Quoted By:

>>12956130
>Remember, not easy for middle ground companies to implement != dead field

Yeah, this is what I was hoping to hear. If I may reveal my ignorance, I am specifically asking about undiscovered, cheaper ways to approximate solutions for the rendering equation. I'd figure everyone might already be on that, though.

Anonymous

Anonymous Sat 10 Apr 2021 18:52:20 No.12956265 Report

Quoted By:

>>12956086
Ultimately intend to complete a graduate degree in computational engineering or data science. I'm already familiar with discrete math, proofs, and linear equations, I just want a good program that will give me a robust and well rounded math/CS centric education so I can be set to have a decent application for a graduate program at a place like John hopkins. I was looking at Rice and their computation program, but it looks pretty shit, I'm not sure why its such a well regarded school.

Anonymous

Anonymous Sat 10 Apr 2021 19:04:23 No.12956334 Report

Quoted By: >>12956425

>>12956100
there's a ton of work being done with integrating ML into graphics, but that's probably not what you're after, and it will be in the saturated S-curve region by the time you can graduate anyway

Anonymous

Anonymous Sat 10 Apr 2021 19:22:59 No.12956425 Report

Quoted By:

>>12956334
Eh, given the discussion about discrete differential geometry last time, and the historical evidence, I think it’s safe to say that graphics is a kind mistress where there’s a lot of theory and applied work to be done.

Anonymous

View Same Google iqdb SauceNAO 717eCUskYwL.jpg, 212KiB, 900x1164

Anonymous Sat 10 Apr 2021 20:01:18 No.12956585 Report

Quoted By: >>12956767 >>12956773 >>12960021 >>12960759

>>12893521
The computer vision push i'm seeing now along with all the shit from the "4th industrial revolution" as just part of the trans-humanist apocalypse which I will fight with my full strength
It's so fucking obvious, all this push to put machines in humans and substitute humans for machines

Anonymous

Anonymous Sat 10 Apr 2021 20:53:02 No.12956767 Report

Quoted By:

>>12956585
Thank you anon, very cool.

Anonymous

Anonymous Sat 10 Apr 2021 20:54:54 No.12956773 Report

Quoted By: >>12956808

>>12956585
The only way to stop it is to destroy technological civilization and that's retarded because even in a cyberpunk tech dystopia life is infinitely better and more fun than in a pre-industrial shithole.

Anonymous

Anonymous Sat 10 Apr 2021 21:01:17 No.12956808 Report

Quoted By: >>12958271 >>12959563 >>12959579

>>12956773
>even in a cyberpunk tech dystopia life is infinitely better and more fun than in a pre-industrial shithole
People did not need glasses in pre-industrial shitholes
People didnt have microplastics in their placenta in pre-industrial shitholes
People didnt have cavities or malformed faces in pre-industrial shitholes
Men's sperm count was going to 0 in pre-industrial shitholes
People didnt need to even consider becoming machines to escape from the hell the same technology created in pre-industrial shitholes

https://www.npr.org/sections/health-shots/2013/02/24/172688806/ancient-chompers-were-healthier-than-ours

>The only way to stop it is to destroy technological civilization
It doesnt need to stop, people just need to seek the truth and follow the Bible

Anonymous

Anonymous Sun 11 Apr 2021 03:39:34 No.12958271 Report

Quoted By: >>12960722

>>12956808
schizoposting

Anonymous

Anonymous Sun 11 Apr 2021 07:36:44 No.12958790 Report

Quoted By:

>>12898645
Too bad ICML has such a big list of accepted papers.

Anonymous

Anonymous Sun 11 Apr 2021 08:38:59 No.12958952 Report

Quoted By: >>12959560

Reminder not to forget to actually build something every now and then. Excessive amounts of theory will also rot your brain.

Anonymous

Anonymous Sun 11 Apr 2021 12:53:51 No.12959560 Report

Quoted By:

>>12958952
Yeah I agree. There are some problems that you only see when you actually implement something.

Anonymous

Anonymous Sun 11 Apr 2021 12:55:46 No.12959563 Report

Quoted By: >>12960028

>>12956808
>People did not need glasses in pre-industrial shitholes
>People didnt have microplastics in their placenta in pre-industrial shitholes
>People didnt have cavities or malformed faces in pre-industrial shitholes
>Men's sperm count was going to 0 in pre-industrial shitholes
>People didnt need to even consider becoming machines to escape from the hell the same technology created in pre-industrial shitholes
Yeah, they just died in childhood, of the plague or starvation instead.

Anonymous

Anonymous Sun 11 Apr 2021 13:03:20 No.12959579 Report

Quoted By: >>12960028

>>12956808
Almost all of that happened in pre industrial societies, plus starvation and disease. Pre industrial civilization wasn't some paradise on earth, there's a reason England had to ban peasants from leaving the fields for the city, industry provided a huge jump in quality of life.

Anonymous

Anonymous Sun 11 Apr 2021 15:30:20 No.12960021 Report

Quoted By: >>12960036

>>12956585
>I will fight with my full strength
how cute

Anonymous

Anonymous Sun 11 Apr 2021 15:31:43 No.12960028 Report

Quoted By: >>12963671

>>12959563
>Yeah, they just died in childhood, of the plague or starvation instead.
They didn't moron, child death might have been high, but survivors could live even until their 100s

>>12959579
>Almost all of that happened in pre industrial societies, plus starvation and disease
Maybe disease but the rest simply did not happen before the tribal people start being introduced to modern's men diet (wheat, sugar, refined flour, coffee, chocolate, etc.)

>Pre industrial civilization wasn't some paradise on earth
I used to think that too but I'm not sure it was any worse than now

> there's a reason England had to ban peasants from leaving the fields for the cit
I'm talking about people that didn't practice agriculture and subsisted off grains, think Weston Price studies

> industry provided a huge jump in quality of life
What do you mean by quality of life? Plastics in placenta? Everyone becoming myopic? Cancer? Vaccines creating more illnesses? Autism? Alzheimer? People not knowing how to make a fork? Trannies? Jews? Big pharma? EMF? Sperm count going to zero in 50% of men in the USA at 2045? Protecting criminals because they're niggers? What fucking quality of life you're talking about?

Anonymous

Anonymous Sun 11 Apr 2021 15:33:12 No.12960036 Report

Quoted By: >>12960722

>>12960021
bet you wont like when i throw salt water on your central servers nerd

Anonymous

Anonymous Sun 11 Apr 2021 18:12:50 No.12960722 Report

Quoted By:

>>12960036
>>12958271

Anonymous

Anonymous Sun 11 Apr 2021 18:22:59 No.12960759 Report

Quoted By: >>12960961

>>12956585
Humanity is disgusting and degenerate. Machines are inherently superior. I strive for the day where the human race ceases to exist.

Anonymous

Anonymous Sun 11 Apr 2021 19:10:02 No.12960925 Report

Quoted By: >>12961017 >>12961175

yoo boys, what the pattern here ?
input : 2 output : 2
input : 5 output : 27
input : 8 output : 248
input : 10 output : 1014

Anonymous

Anonymous Sun 11 Apr 2021 19:17:02 No.12960961 Report

Quoted By:

>>12960759
>Humanity is disgusting and degenerate
that's a lot of projection buddy

>I strive for the day where the human race ceases to exist
imagining striving to become a being that needs constant software updates and eventually crashes because of hardware issues that can't be fixed by your own body
hey man if you want the troooonshumanist experience go take a vaccine every week

Anonymous

Anonymous Sun 11 Apr 2021 19:25:50 No.12961017 Report

Quoted By: >>12963609

>>12960925
2^x-x

Anonymous

Anonymous Sun 11 Apr 2021 19:53:33 No.12961175 Report

Quoted By: >>12963609

>>12960925
it can be literally anything
sequences are meaningless

Anonymous

Anonymous Sun 11 Apr 2021 19:57:57 No.12961190 Report

Quoted By: >>12962562

has research pertaining to logics been mostly replaced with research pertaining to programming languages?
moreoever, at this point, are there any ways to actually improve programming languages besides syntactic sugar?

Anonymous

Anonymous Mon 12 Apr 2021 02:19:00 No.12962562 Report

Quoted By:

>>12961190
>are there any ways to actually improve programming languages besides syntactic sugar?
read PLT research
the answer is yes

Anonymous

Anonymous Mon 12 Apr 2021 02:56:52 No.12962659 Report

Quoted By: >>12962959

>>12893521
Computer Vision: Algorithms and Applications (free online!):
https://szeliski.org/Book/
The bottom of the page has multiple courses and slide sets on the same material, provided both by this instructor and many others.

Stanford CS 131 CV class (2018):
http://vision.stanford.edu/teaching/cs131_fall1819/syllabus.html
CS 131 (2014):
http://vision.stanford.edu/teaching/cs131_fall1415/index.html

Stanford 231 CV (harder than 131) lecture videos:
https://www.youtube.com/playlist?list=PLf7L7Kg8_FNxHATtLwDceyh72QQL9pvpQ

Are these the best resources to get into Computer Vision?

Anonymous

Anonymous Mon 12 Apr 2021 04:50:47 No.12962959 Report

Quoted By: >>12962975

>>12962659
Do you have any better? These are some good resources recommended to me by friends and some ones I've used from time to time.

Anonymous

Anonymous Mon 12 Apr 2021 04:56:23 No.12962975 Report

Quoted By:

>>12962959
Nope I have none to recommend. Just asking

Anonymous

Anonymous Mon 12 Apr 2021 10:01:51 No.12963609 Report

Quoted By: >>12963669

>>12961175
See >>12961017

Anonymous

Anonymous Mon 12 Apr 2021 10:38:14 No.12963669 Report

Quoted By: >>12966890

>>12963609
That's one curve that fits your points. There are literally infinitely many others. You didn't specify any constraints for the function.

Anonymous

Anonymous Mon 12 Apr 2021 10:39:16 No.12963671 Report

Quoted By:

>>12960028
>They didn't moron, child death might have been high, but survivors could live even until their 100s
Same today except child death is low, dipshit. And quality of life is better.

Anonymous

Anonymous Mon 12 Apr 2021 11:12:59 No.12963739 Report

Quoted By: >>12964153 >>12964298 >>12965696 >>12965780

How do you guys balance self-studying math-heavy topics? I want to get to the meaty topics in TCS fast but it is taking me ages kek.

Anonymous

Anonymous Mon 12 Apr 2021 13:57:05 No.12964153 Report

Quoted By:

>>12963739
I bookmark some books then go back to watching anime

Anonymous

Anonymous Mon 12 Apr 2021 14:47:20 No.12964298 Report

Quoted By:

>>12963739
Pick one topic and do it for a few months. Be sure to do problems.
I don’t allow myself to pick up another book unless
1) I hate the current one
2) I finish at least 4 chapters of the current book

Anonymous

Anonymous Mon 12 Apr 2021 20:01:55 No.12965696 Report

Quoted By:

>>12963739
do it on a need to know basis
though if you want to get into something like geometric complexity, you're gonna need a lot of background knowledge lmao

Anonymous

Anonymous Mon 12 Apr 2021 20:17:12 No.12965780 Report

Quoted By:

>>12963739
watch porn and edge my cock when reading/doing math so I get a pavlov's reaction to it and do it more

Anonymous

Anonymous Mon 12 Apr 2021 20:36:21 No.12965884 Report

Quoted By: >>12969133 >>12969330

is "algorithmic bias" (or bias in AI) a serious issue or is it something that only receives a lot of attention these days for political reasons?
this is a genuine question btw.

Anonymous

View Same Google iqdb SauceNAO 1602962359010.png, 100KiB, 959x344

Anonymous Mon 12 Apr 2021 22:56:39 No.12966546 Report

Quoted By:

So I'm studying heuristics but I'm kind of unsure about the definition of a "perfect heuristic".

Take the initial state considered in picrel. If I choose the Manhattan distance of each box to the closest objective as a heuristic, we have a heuristic value of 10, correct?
Because it's admissable, we have that h(so) <= h*(so). But what the fuck *is* the perfect heuristic h*? In this case, is it the same as the Manhattan distance?

Anonymous

Anonymous Tue 13 Apr 2021 00:20:35 No.12966890 Report

Quoted By: >>12967232

>>12963669
if OP didn't specify any other points, then there's no reason for this to be wrong.

Anonymous

Anonymous Tue 13 Apr 2021 01:53:41 No.12967232 Report

Quoted By: >>12967278

>>12966890
I never said it was wrong. But there is no correct answer either.

Anonymous

Anonymous Tue 13 Apr 2021 02:00:14 No.12967256 Report

Quoted By: >>12967287 >>12969157

>>12893521
Serious question, I turn into a literal retard whenever complex math comes up in CS and I struggle when trying to learn CV and ML. Are the links in OP the best resources for retards?

Anonymous

Anonymous Tue 13 Apr 2021 02:08:06 No.12967278 Report

Quoted By: >>12969190

>>12967232
It is. When you force the problem to have a closed form, then there is an actual solution. Read more on generating functions

Anonymous

Anonymous Tue 13 Apr 2021 02:09:43 No.12967287 Report

Quoted By:

>>12967256
What do you mean by complex or by retard? At some point there’s no point to dumbing down math, CS or otherwise.
Lemme say it another way: why exactly do you think the complex math in CS is complex? Do you think it’s because it’s dickwaving, or do you think it’s because the problem is complex?

Anonymous

Anonymous Tue 13 Apr 2021 14:22:58 No.12969133 Report

Quoted By: >>12969317

>>12965884
If you are talking about the inaccuracies when it comes to classification of black people photos, and the lack of female representation in ML translation, then that can be fixed with better control on data and has nothing to do with the algorithms themselves.

Anonymous

Anonymous Tue 13 Apr 2021 14:29:29 No.12969157 Report

Quoted By: >>12969327 >>12969530 >>12972012

>>12967256
I got a 2.2/4 for my Calculus 2 class and 2.6 for my Linear Alg class, failed stat once.
I still read and understand ML papers now. The truth is I hated the way these topics were taught in 1st/2nd year in college. I always feel so put off by them.
Later, I realized I had to relearn them to do ML (I like ML). Once I put away the annoyance feeling toward them I learn by myself just fine.

Anonymous

Anonymous Tue 13 Apr 2021 14:43:44 No.12969190 Report

Quoted By:

>>12967278
There are infinitely many closed form solutions to drawing a curve between points. You can start with polynomials of degree n but that's just one family of functions that can be used. Plus the original problem said nothing about the form of solution. You could just define a simple mapping of points and it would be equally valid.

Anonymous

Anonymous Tue 13 Apr 2021 15:13:21 No.12969317 Report

Quoted By:

>>12969133
that's kind of what I thought
for some dumb assignment we had to implement some adversarial learning models that would "combat" algorithmic bias (for instance, neutralise the sentiment of word embeddings for demographic identity terms). so i had to read a bunch of research papers into this whole area.
the whole time I was thinking how stupid the term "algorithmic bias" is because the source of all of these so called biases was the data, not the algorithms.
anyway, I'm glad I don't have to waste my time with that shit anymore, it's incredibly dull

Anonymous

Anonymous Tue 13 Apr 2021 15:15:15 No.12969327 Report

Quoted By: >>12969530

>>12969157
to be honest this sounds like a similar experience to my own. I was always good at maths and always liked it, but I got more *into* maths after doing straight CS, beause I enjoyed studying the mathematical concepts that underlied everything aside my course, rather than having it imposed onto me.

Anonymous

Anonymous Tue 13 Apr 2021 15:15:41 No.12969330 Report

Quoted By:

>>12965884
Normaflags cannot into GIGO and demand IT do something about it

Tale as old as time.

Anonymous

Anonymous Tue 13 Apr 2021 16:06:31 No.12969530 Report

Quoted By: >>12970547

>>12969157
>>12969327
They try and sanitize the pedagogy early on, so you have no idea what calculus and linear algebra are good for, especially in CS.
That, and the fact that calc 2 is the first STEM weed out course, it's not uncommon to see a lot of CS majors appreciate math late into their career. I think this is why it's really important for CS departments to actually motivate the theory they teach.

I want to laugh at many CS majors for not being able to do basic calculus, but the sad truth is that a lot of CS departments don't do enough to show how calculus is used in things *as mundane as* analyzing an algorithm or solving some classic problems from other fields of science using computers in a clever way. Unironically, I think if they added a class called "theoretical tools and their applications" to the CS canon, where they taught hardline uses of generating functions, correlation inequalities, spectral graph theory, vector calculus, and Fourier transforms (hell, those are even used in the discrete boolean setting), I think CS majors might *groan* but would be unable to say the "yawn, where is this useful?" That, and they'd scoff less at the calculus because it actually matters.

Anonymous

Anonymous Tue 13 Apr 2021 20:03:56 No.12970547 Report

Quoted By: >>12970925 >>12974647 >>12974845

>>12969530
Yeah what you said is true. But a professor in my dept told me that there is a real pressure from industry on schools to make CS courses focus more on "code" and "problem solving".
And a lot of the students want to code things like games and software don't help.
Making a mandatory theory course and their applications in CS would be good but there has to be someone who is a big deal in the field to start it first.

Anonymous

Anonymous Tue 13 Apr 2021 21:23:25 No.12970925 Report

Quoted By:

>>12970547
Yeah there seems to have been lots of pressure from the industry to dumb down CS programs huh.

Anonymous

Anonymous Wed 14 Apr 2021 01:26:53 No.12972012 Report

Quoted By: >>12972732

>>12969157
what resources were the most useful in learning ML and the relevant math?

Anonymous

Anonymous Wed 14 Apr 2021 06:17:27 No.12972732 Report

Quoted By:

>>12972012
Not the guy you’re responding to, but it really depends on what part of ML you want to work with. But everyone eventually has to hunker down to understand Hilbert space, measure theory, etc.

Anonymous

Anonymous Wed 14 Apr 2021 16:55:10 No.12974647 Report

Quoted By: >>12974877

>>12970547
Lmao if only they taught CS majors at the less better schools how to solve problems though. Even ABET CS standards dip mostly into codemonkeying, to the point where Berkeley doesn’t even follow them for EECS anymore.

Anonymous

Anonymous Wed 14 Apr 2021 17:28:33 No.12974845 Report

Quoted By: >>12974877 >>12974953

>>12970547
It won't help because there will be shitty schools with retarded profs who know nothing and teach shit. It applies both to engineering and theoretical part of CS or Math. If only there would be a solution, like a certified, unified school with 100% employment and internship chance that is accessible for everyone. Good teachers and professionals could create a curriculum and the bad ones could be simple instructors following the program, easy as that.

And you can answer any "Why i need to know that, uggh" with a solid "Cuz you're getting a JOB, nigga!"

Anonymous

Anonymous Wed 14 Apr 2021 17:35:14 No.12974877 Report

Quoted By: >>12974901 >>12975055

>>12974845
>>12974647
It's hilarious how little math CS majors from average schools actually know. Used to think CS wad a math heavy course but when I hopped over to CS/CE after my BEng I was shocked to see half the class get stumped by fucking Fourier series in the networking class. Got top scores across the board in math subjects despite having a brainlet engineer background.

Anonymous

Anonymous Wed 14 Apr 2021 17:39:47 No.12974901 Report

Quoted By: >>12974960

>>12974877
it’s hilarious how much dunning kruger engineers from state schools experience bc they took an extra diffy qs / calc 3 class.

Anonymous

Anonymous Wed 14 Apr 2021 17:48:43 No.12974945 Report

Quoted By:

anyone else feels anxious about so many people going into cs and discussing how to get better chances? i dont like competition

Anonymous

Anonymous Wed 14 Apr 2021 17:51:42 No.12974953 Report

Quoted By:

>>12974845
>If only there would be a solution, like a certified, unified school with 100% employment and internship chance that is accessible for everyone
you retards should be taught basic economics as well because there is no way to employ 100% of people for nothing, value must be generated

your rewards are in exact porportion to how hard it is to substitute you, how hard it is to do what you do and how well you do it

Anonymous

Anonymous Wed 14 Apr 2021 17:52:40 No.12974960 Report

Quoted By: >>12975055

>>12974901
Never said I am smarter than you dumbass. I know I am an engineer from a shitty euro university. Just pointing out that the level is so low that even engineers can dunk on the average CS grad.
>extra diffy qs
For what it's worth I'm doing better across the board, including clases like number, algorithm, graph theory, crypto, probability and stats, etc. Not to mention any class which involves practical programming.

I think it's more a difference in acquired work ethic and bullshit tolerance between a 3 year academic degree and a 5 year professional one.

Anonymous

Anonymous Wed 14 Apr 2021 18:19:35 No.12975055 Report

Quoted By:

>>12974877
>>12974960
It really depends on your program. That’s the unfortunate truth with CS - it’s a great field whose undergrad doesn’t adequately prepare you for the breadth and depth of the things you might study.
Huge difference between the crowd doing work on programmable metamaterials (both computer science and mechanical engineering):
https://youtu.be/BTHJuVoH6Ro
And your average person who wants to do business logic as a dev.

Anonymous

Anonymous Thu 15 Apr 2021 05:50:56 No.12977455 Report

Quoted By: >>12977623

New thread?

Anonymous

Anonymous Thu 15 Apr 2021 07:12:29 No.12977623 Report

Quoted By:

>>12977455
Yeah I’ll put it up soon.

Anonymous

View Same Google iqdb SauceNAO ugh-pepe.png, 469KiB, 1099x1012

Anonymous Thu 15 Apr 2021 07:27:56 No.12977656 Report

Quoted By:

Seems like this is the right place to ask.
I am in high school, applying to Dutch unis.
Now I have the option to chose between BSc Artificial Intelligence in University of Gronigen and BSc Computer Science and Engineering in Delft University of Technology.
What would you do in my situation and why?

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

Your latest searches