/sci/ - Science & Math » Thread #12793853

213KiB, 1497x499, imgonline-com-ua-twotoone-vSvIQ2XFqm.jpg

View Same Google iqdb SauceNAO

/cspg/ Computer Science Plus General

Anonymous Fri 05 Mar 19:37:37 2021 No.12793853 View Reply Original Report

Quoted By: >>12793989 >>12810523 >>12811030 >>12812604 >>12813023 >>12813396 >>12820097 >>12828088

Learn programming and theory from scratch with strong fundamentals. /cspg/ guides and textbook packs coming soon!

Topics: CS + SWE + CE + IT (A+ certs, etc) + SecOps + Math

Guides:
Complete self taught pathway: https://teachyourselfcs.com/
Structured pathway: https://github.com/ossu/computer-science
AI focused pathway (warning: functional programming ahead): https://learnaifromscratch.github.io/

The first link recommends SICP as an entry, but if you want a gentler introduction to the world of computing, I'd recommend How to Design Programs. Before you complain, read:
http://cs.brown.edu/~sk/Publications/Papers/Published/fffk-htdp-vs-sicp-journal/

Math for CS:
https://stackoverflow.com/questions/14868480/math-for-computer-science
https://4chan-science.fandom.com/wiki/Mathematics

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

Not sure if programming is for you? Try:
https://1lib.us/book/11284291/ccb571 (New edition)

New study groups for March 15th confirmed for:
- Structure and Interpretation of Computer Programs (SICP)
- How to Design Programs (HtDP)
- Programming and Programming Languages (PAPL)
- Introduction to Algorithms by CLRS (CLRS)
- Concrete Mathematics: A Foundation for Computer Science by Knuth et al
- Modern Compiler in ML by Appel (potentially)
- Others being considered

Discord: https://discord.gg/S2j9jJJGKM

Remember, you don't need to be a computer scientist to be a good programmer, but it sure does help.

Anonymous

Anonymous Fri 05 Mar 2021 19:42:31 No.12793879 Report

Quoted By: >>12794729 >>12796114 >>12796840 >>12821051 >>12823367 >>12823371 >>12823377

Anyone think the ML bubble is gong to pop?
Or they will last for a few decades?
Apart from computer vision, a lot of the research papers have reproducibility problems.

Anonymous

Anonymous Fri 05 Mar 2021 19:48:15 No.12793914 Report

Quoted By: >>12794034 >>12808573 >>12813367

where should i start from as an absolute complete noob to learn computer science?

Anonymous

Anonymous Fri 05 Mar 2021 20:02:27 No.12793989 Report

Quoted By: >>12796465

>>12793853
I want to NEET and try to build a start up after my degree instead of jumping into industry.
Does industry care?

Anonymous

Anonymous Fri 05 Mar 2021 20:10:58 No.12794034 Report

Quoted By:

>>12793914
Kindergarden

Anonymous

Anonymous Fri 05 Mar 2021 20:35:28 No.12794146 Report

Quoted By: >>12794163 >>12796450

Is computer science thread always thia dead?

Anonymous

Anonymous Fri 05 Mar 2021 20:38:30 No.12794163 Report

Quoted By: >>12794392 >>12794751 >>12795831 >>12813076 >>12823046

>>12794146
CS isn't considered a real field by /sci/ standard.
Too little hard engineering for eng, too little math for math, and too little real lab work for science.
The funny thing is half of the math people probably need to learn how to code to make money anyway.

Anonymous

Anonymous Fri 05 Mar 2021 21:19:22 No.12794392 Report

Quoted By: >>12795156 >>12796114

>>12794163
>too little math
name one field of study (other than math itself, and theoretical physics) that uses more.

Anonymous

Anonymous Fri 05 Mar 2021 22:28:59 No.12794729 Report

Quoted By:

>>12793879
ML is just a fancy name for the latest in CS research, just like AI which was used in the 70s for stuff that is no longer considered AI but was the current frontier of research. it took decades and decades to arrive at what we currently call ML and there is still a lot of work to be done, so ML will stick around for a while.

Anonymous

Anonymous Fri 05 Mar 2021 22:34:17 No.12794751 Report

Quoted By: >>12794777

>>12794163
>Too little hard engineering for eng, too little math for math, and too little real lab work for science.
this is just all wrong. your impression of CS seems to be codemonkey business. Theoretical CS work is just math, and you don't need lab work to do science. labs are for experimental science, which is the most soul draining science.

sadly very few people seem interested in discussing the actual science. this seems to be universal on 4chan though, even /mg/ is shit tier math and usually not even math.

Anonymous

Anonymous Fri 05 Mar 2021 22:41:13 No.12794777 Report

Quoted By:

>>12794751
>your impression of CS seems to be codemonkey business.
That's not my impression. I was just interpreting the general consensus of /sci/ toward CS and the reason why threads like these don't have many people.
I'm a PhD student in theoretical CS myself and I hate that too but I can't change what the majority /sci/ thinks.
But I could be wrong and maybe most CS undergrads are in /g/ because it's more active over there so you can get your answers faster.

Anonymous

Anonymous Sat 06 Mar 2021 00:25:46 No.12795156 Report

Quoted By:

>>12794392
Mechanics (classical and solid)

Anonymous

Anonymous Sat 06 Mar 2021 01:34:47 No.12795324 Report

Quoted By: >>12795831 >>12797153

Maybe people can post their homework to start discussion I don't know.

Anonymous

View Same Google iqdb SauceNAO het_bdpt_transient_64_false (1).webm, 250KiB, 512x512

Anonymous Sat 06 Mar 2021 04:03:52 No.12795831 Report

Quoted By:

>>12794163
Unfortunately, this is the image CS has because most people who do CS undergrad just want enough to do a codemonkey career, and this has been really bad PR for CS in the eyes of young STEM undergrads. As much as you tell them it's not the case, they'll always deny it because some person's out in the tech world writing code without being able to do a simple proof or work on anything more complex than a CRUD app
>>12795324
Here's an interesting recent project from CMU's computer graphics lab:
https://imaging.cs.cmu.edu/rrte/
Abbreviated abstract:
"We propose several acceleration techniques to handle complex scenes (surfaces and volumes) that include participating media with heterogeneous refractive fields. We use our algorithms to simulate a variety of scenes combining heterogeneous refraction and scattering, as well as tissue imaging techniques based on ultrasonic virtual waveguides and lenses. Our algorithms and publicly-available implementation can be used to characterize imaging systems such as refractive index microscopy, schlieren imaging, and acousto-optic imaging, and can facilitate the development of inverse rendering techniques for related applications."

Video related

Anonymous

Anonymous Sat 06 Mar 2021 04:27:54 No.12795896 Report

Quoted By: >>12795998

I need to understand the PageRank and eventually the TextRank algorithm. Any resources?

Anonymous

Anonymous Sat 06 Mar 2021 04:54:46 No.12795998 Report

Quoted By: >>12796014

>>12795896
https://cs50.harvard.edu/ai/2020/projects/2/pagerank/

It's overrated:
>The key to real-time search, Norvig said, is Google's famously distributed back-end infrastructure, which is able to re-build its web index with relatively little delay.
https://www.theregister.com/2010/03/03/google_research_head_norvig_on_pagerank/ old clickbait but still true, it's the infrastructure of these outfits that excel with seach not any algorithm.

Anonymous

Anonymous Sat 06 Mar 2021 04:59:20 No.12796014 Report

Quoted By:

>>12795998
Thanks anon. I do not care about Pagerank specifically, just need to implement Textrank (which is based on pagerank)

Anonymous

View Same Google iqdb SauceNAO sleep sort.jpg, 157KiB, 1020x732

Anonymous Sat 06 Mar 2021 05:23:19 No.12796098 Report

Quoted By: >>12800192 >>12800216 >>12807608

What would this kind of sort be classified as?
In an ideal scenario in which the CPU can differentiates all timestamps in the array, it has running time O(max(arr)) right?

Anonymous

Anonymous Sat 06 Mar 2021 05:26:59 No.12796107 Report

Quoted By: >>12796119 >>12821057

what kind of actual research is going in on distributed systems?
are there problems that actually could be applied to high performance shit?

i'm considering going back to school for a masters or phd in distributed systems, but i'd rather not become a soulless npc doing absolutely nothing with my life except making jeff, bill, or zucc richer

Anonymous

Anonymous Sat 06 Mar 2021 05:28:09 No.12796114 Report

Quoted By: >>12796154 >>12796256 >>12796466

>>12793879
depending on who you ask the bubble is already over

>>12794392
every other engineering field, at the undergrad level

Anonymous

Anonymous Sat 06 Mar 2021 05:30:14 No.12796119 Report

Quoted By: >>12796138

>>12796107
>what kind of actual research is going in on distributed systems?
There are a lot actually. I usually go to an algorithm seminar in my school, and from what I gathered, many algorithm / data structure can be improved with distributed systems.
The field is difficult so there's always a demand for people.

Anonymous

Anonymous Sat 06 Mar 2021 05:35:32 No.12796138 Report

Quoted By:

>>12796119
i'm currently working, i'd want to work on research that improovs the performance of hpc compute and shit. the main reason i brought up distributed systems because i have the most opportunities to get industrial exposure to this field

Anonymous

Anonymous Sat 06 Mar 2021 05:43:47 No.12796154 Report

Quoted By: >>12797505

>>12796114
>at the undergrad level
who gives a shit

Anonymous

Anonymous Sat 06 Mar 2021 06:32:21 No.12796256 Report

Quoted By: >>12797909

>>12796114
>more math
CS majors might be shitters at other schools, but learning how to massage basic differential equations and how to use lookup table does not constitute “more math.”

Anonymous

Anonymous Sat 06 Mar 2021 07:33:35 No.12796378 Report

Quoted By: >>12796379 >>12797188

I know a lot of CS students lurk /sci/ (based on some past surveys) but it feels like no one gives a shit about this thread.

Anonymous

Anonymous Sat 06 Mar 2021 07:34:28 No.12796379 Report

Quoted By: >>12797493

>>12796378
CS students don't care about anything besides passing and getting a job after graduating.

Anonymous

Anonymous Sat 06 Mar 2021 08:05:50 No.12796450 Report

Quoted By: >>12797493

>>12794146
cs isn't a real science. Anyone in cs is just a neet that wants to make money without breaking a sweat.

Anonymous

Anonymous Sat 06 Mar 2021 08:09:13 No.12796465 Report

Quoted By:

>>12793989
yep, unless you make a killer portfolio in the process

Anonymous

Anonymous Sat 06 Mar 2021 08:09:16 No.12796466 Report

Quoted By: >>12797505 >>12813085

>>12796114
>every other engineering field
How delusional are you?

Anonymous

View Same Google iqdb SauceNAO 1615019616339.png, 315KiB, 600x412

Anonymous Sat 06 Mar 2021 10:35:16 No.12796840 Report

Quoted By: >>12820138

>>12793879
As long as we can throw MOAR COMPUTE LMAO at the problem ML will just keep moving forward.

Anonymous

Anonymous Sat 06 Mar 2021 13:39:22 No.12797153 Report

Quoted By:

>>12795324
lol sure
I'm doing some ML coursework, we're basically supposed to do something with this coronavirus dataset: https://github.com/beoutbreakprepared/nCoV2019/tree/master/latest_data

The dataset has over 2 mil columns, but for the "outcome" column (i.e., recovered from covid or died) there are only 300,000 non-null values. That seems fine, okay a dataset of 300k is still pretty substantial, but then the only other features that contain mostly non-null values for this subset are longitude and latitude and date confirmation, and country. Only a subset of about 40k or less here has non-null values for age, which is probably the most important predictor. So if i reduce the dataset down again I'm now working with a dataset with 40,000 rows, as compared to the original with 2 mil rows. And the worst thing is these instances are almost entirely concentrated in India and the Phillipines of all places. kinda frustrating desu, we've been given such a crappy dataset for which only 2% is usuable.

Anonymous

Anonymous Sat 06 Mar 2021 13:40:38 No.12797154 Report

Quoted By:

got 98% on my last summative lads

Anonymous

View Same Google iqdb SauceNAO file.png, 64KiB, 635x518

Anonymous Sat 06 Mar 2021 13:44:50 No.12797164 Report

Quoted By:

thoughts on the year 3 modules i can choose?
bearing in mind I have a max of 120 credits, and all of them apart from individual project (40 creds) and codes and cryptography (20 creds) are 10 creds.
how does it compare to your uni's? guessing this is a kinda standard array of choices

Anonymous

Anonymous Sat 06 Mar 2021 14:00:54 No.12797188 Report

Quoted By: >>12797493

>>12796378
They only care about doing leetcode and getting that FAGMAN job

Anonymous

View Same Google iqdb SauceNAO AFED55C9-5CED-4ECC-82E4-EC4139CB (...).jpg, 63KiB, 556x544

Anonymous Sat 06 Mar 2021 15:59:03 No.12797493 Report

Quoted By:

>>12796379
>>12796450
>>12797188

Anonymous

Anonymous Sat 06 Mar 2021 16:02:51 No.12797505 Report

Quoted By: >>12797532 >>12797885

>>12796154
most people who only do undergrad then leave

>>12796466
engineering dropout who went into CS undergrad. engineering math was much harder imo

Anonymous

Anonymous Sat 06 Mar 2021 16:11:35 No.12797532 Report

Quoted By: >>12797552 >>12797918

>>12797505
>most people who only do undergrad then leave
I was strictly talking about serious CS and research output. you’re just trying to make a point that a lot of modern undergraduate programs are degree mills. no one is denying this but it’s not relevant to the point I was trying to make.
>>12797505
>engineering dropout who went into CS undergrad. engineering math was much harder imo
as an engineer I doubt you touched any serious math.

Anonymous

Anonymous Sat 06 Mar 2021 16:18:34 No.12797552 Report

Quoted By:

>>12797532
>no one is denying this but it’s not relevant to the point I was trying to make.
gotta be distinct. almost all of the criticism towards cs is directed towards the undergrads because we do feel like we're hot shit because we get handed six figures out of that diploma mill

>as an engineer I doubt you touched any serious math.
never claimed i did

Anonymous

Anonymous Sat 06 Mar 2021 18:08:47 No.12797885 Report

Quoted By: >>12797978

>>12797505
>dropout who went into CS
I hear this a lot from undergrads in EE. Given my experience with both CS/CSE and mechE, I think it's less that "CS is unquestionably easier" as much as it's that at most schools, the skill floor for CS undergrad is lower. That's an effect of it being new and because most of the real reasons why it matters is locked behind the upper div classes and grad classes where you obviously need more math/rigor/etc. to do well.

So you get a lot of EE dropouts who did the bare minimum in CS to get a piece of paper, versus the EE program where its skill ceiling is lower but most consistent since everyone has to do relatively the same program, save like 3-4 classes. I could see how this could color everyone's perception of CS as the "safe, easier" alternative. Obviously CS isn't easy at a good school, doesn't have to do with code, etc.., but all of that gets buried under
>broo I dropped out of EE and CS is piss easy!
>broo I just chose this because I want a job without trying
>broo what do you mean the Fast Fourier Transform is used to multiply polynomials quickly? I can just use a linked list I learned in my intro data structures class
Lots of CS undergrads suck, but it isn't fair that CS is seen as the brainlet major. It'll probably rectify itself as time goes on, standards get higher, the subject's foundations mature, and CS is seen less and less about being about code and programming.

Anonymous

Anonymous Sat 06 Mar 2021 18:15:25 No.12797909 Report

Quoted By: >>12797919 >>12797999

>>12796256
cope & seethe faggot, most engis learn pde's, complex analysis, numerical analysis among others

Anonymous

Anonymous Sat 06 Mar 2021 18:17:48 No.12797918 Report

Quoted By: >>12797947 >>12797969 >>12798026

>>12797532
>as an engineer I doubt you touched any serious math
Wow guys i took a course on discrete math thats similar to what a math major takes wow guys im doing real math!

Anonymous

Anonymous Sat 06 Mar 2021 18:17:56 No.12797919 Report

Quoted By:

>>12797909
>complex analysis
this isn’t true unless you’re an EE.
>pdes
rote memorization.

Anonymous

Anonymous Sat 06 Mar 2021 18:22:44 No.12797947 Report

Quoted By:

>>12797918
I double majored in math retard. and again, in my original post I was talking about the field of CS relative to others.

Anonymous

Anonymous Sat 06 Mar 2021 18:26:29 No.12797969 Report

Quoted By:

>>12797918
also didn’t take discrete math. I took combinatorics instead to meet the requirement.

Anonymous

Anonymous Sat 06 Mar 2021 18:29:44 No.12797978 Report

Quoted By: >>12797995

>>12797885
yeah more or less i agree wtih you. the vast majority of people doing a BS or MS are solely in it to become SWEs. the prevailing mindset in CS is "this class is worthless because it won't get me a job at google" and the amount of people focused on industry, which usually requires little to no math at all, vastly outnumbers those interested in research

a lot of csfags see other fields get completely fucked in the job market, so getting a job is #1 priority.
field has a way of ruining your self esteem. i'm considered below average because i only make 150k at age 25

>CS is seen less and less about being about code and programming.
this will happen when the tech bubble bursts and SWE stops being the most desired job in the entire country

Anonymous

Anonymous Sat 06 Mar 2021 18:33:51 No.12797995 Report

Quoted By: >>12798010

>>12797978
>tech bubble bursts
I don’t see this happening anytime soon

Anonymous

Anonymous Sat 06 Mar 2021 18:34:49 No.12797999 Report

Quoted By:

>>12797909
> most engis learn pde's
literally as I said before: massaging basic differentials. PDEs for engineering students is literally just doing laplace transforms to reduce some diff eqs to highschool algebra, like 2 instances of sturm-louiville, and then writing out a heat equation with an infinite series. Not hard.
>complex analysis
Lmao no. Engineering students take a complex variables classes with is baby's first contour integration and calculus with j in the exponential. No complex analysis is used in EE properly until graduate school...which is about the time it shows up properly in CS as well lmao:
http://algo.inria.fr/flajolet/Publications/AnaCombi/book.pdf
>numerical analysis
engineering students take a "numerical methods" class where they learn how to apply well known methods to approximating annoying problems. Most numerical analysis classes I've seen usually live cross listed in math and in CS departments, where there is a greater focus on understanding why and how they work, and implementing them. Engineering students do this to some capacity too, but most of their class time is more spent on matching methods to well known classes problems rather than studying the methods or the math behind them as a focus. Of course you don't learn how to make new ones, since that's nontrivial at this point.

Again, CS is full of shitters and many programs are bad, but engineering "math" is not much more complicated in undergrad, if at all lmao. It's all calculations to make the physics and some optimization problems work out, not actual mathematics.

Anonymous

Anonymous Sat 06 Mar 2021 18:38:34 No.12798010 Report

Quoted By:

>>12797995
People have done my job for half my pay years ago. There is no reason besides the scale of these companies that we're overpaid this much. There's a lot of incentive to cut our salaries and hire less.

Anonymous

Anonymous Sat 06 Mar 2021 18:41:49 No.12798026 Report

Quoted By: >>12817129

>>12797918
>Wow guys i took a course on discrete math thats similar to what a math major takes wow guys im doing real math!
I double majored in math and CS. What CS majors do in discrete math, analysis of algorithms I and II, computability, complexity, programming language theory, and formal verification to name a few is in fact much closer to the math that math majors do than the "math" (aka calculator shit and calculus) engineering students do. CS students aren't as good (across the board, but there are more than a decent amount of bright CS students who'd do fine in math or engineering too) since not everyone there is there for theory, so they do indeed dumb it down more times than not for the required classes and earlier classes like algo I.
But if you're in advanced algo / algo II, randomized algorithms or probabilistic algos, complexity theory, etc.. they have no reason to dumb it down anymore, and you notice these classes are a lot harder. This isn't even mentioning stuff like ML theory bio inspired computing, where both calculus and the pure math come to bite back. I took analysis II (integration theory and introduction to measure) and algo II the same semester, and I can't really tell you which one was harder. Granted, my school is more than decent for both, so YMMV based on what program you're in

Anonymous

Anonymous Sat 06 Mar 2021 18:42:09 No.12798030 Report

Quoted By: >>12798054 >>12817132

I'm in a weird spot, /cspg/. I got used to writing one-off python scripts in grad school for stuff like machine learning models, so I'm not worried about my programming fundamentals. But I've never had to understand the broader world of software engineering, and I'm starting to move in that direction at my company. Topics like APIs, connecting to servers, structure of large projects, all that stuff that isn't just a self-contained .py file on my local drive essentially. Anyone know of resources/advice to help with this kind of situation?

Anonymous

Anonymous Sat 06 Mar 2021 18:48:43 No.12798054 Report

Quoted By:

>>12798030
I don't know much about intermediate resources, but I think the minutia of writing good code that is meant to be reviewed and play nicely with other parts of the project is something you just have to make conscious note of. It'll start getting easier and easier as you keep going forward.

Another way is to do some large personal project that's sensitive to a lot of moving parts. I think this is a bit less effective since industry experience is the most valuable thing you can have to writing industry standard code.

Anonymous

Anonymous Sun 07 Mar 2021 00:50:00 No.12799852 Report

Quoted By:

Anonymous

View Same Google iqdb SauceNAO imagen_2021-03-06_201557.png, 82KiB, 315x315

Anonymous Sun 07 Mar 2021 02:16:03 No.12800192 Report

Quoted By: >>12800193 >>12807839 >>12808522

>>12796098
wait a second?????
isn't sleepsort an algorithm that works in O(n) time?

Anonymous

Anonymous Sun 07 Mar 2021 02:17:17 No.12800193 Report

Quoted By: >>12800287 >>12800347 >>12808522

>>12800192
never mind it's quadratic

Anonymous

Anonymous Sun 07 Mar 2021 02:27:26 No.12800216 Report

Quoted By:

>>12796098
OS is sorting somewhere, the cost is just hidden and it will evidently not be O(n) if you experiement with large enough n. you make any algorithm take what looks like O(n) real time for small enough n. Let n0 be an n that is much larger than any input you expect, and find the real time t0 that it takes to compute the solution for n=n0. then in your "O(n)" algorithm, just use the real algorithm to compute the solution, which takes t time, and then wait n*t0-t time.

Anonymous

Anonymous Sun 07 Mar 2021 02:47:30 No.12800287 Report

Quoted By:

>>12800193
you remind me that technically could be an O(n) "span" algorithm if you analyze it as a parallel algorithm and you distinguish that we are measuring real time instead of the usual which is number of operations. If it takes O(1) real time to start a process and go to the next element in the input, then it takes O(n) real time to start all the processes and no process takes more than O(n) real time to complete, so the total real time is O(n).

I don't know if this is how you got quadratic time, but in parallel algorithms, you could say the real time "work" of this algorithm would be quadratic since there are O(n) processes each taking O(n) real time to complete, and n*n=n^2.

Anonymous

Anonymous Sun 07 Mar 2021 02:58:54 No.12800347 Report

Quoted By: >>12807839

>>12800193
O(max(arr)/gcd(arr)+n*log(n))
assuming the scheduler uses a heap and you optimize the time scale to fit the data

Anonymous

View Same Google iqdb SauceNAO figure2.jpg, 60KiB, 899x503

Anonymous Sun 07 Mar 2021 04:10:24 No.12800570 Report

Quoted By: >>12800965 >>12833953

Keenan Crane has a lot of interesting work using differential geometry in CS, where he notes that the geometry from continuous structures yields theorems in the discrete and geometry from the discrete structures yields theorems in the continuous, etc etc.
His graphics work is really cool. Here's one of the courses he's teaching now:
https://brickisland.net/DDGSpring2021/

Last year's notes on it:
https://brickisland.net/DDGSpring2020/

Crane's own presentation on discrete differential geometry:
https://www.youtube.com/watch?v=KiekJeTEQ-U&t=99s&ab_channel=KeenanCrane

CMU graphics related courses:
http://graphics.cs.cmu.edu/?page_id=16

Anonymous

Anonymous Sun 07 Mar 2021 05:09:15 No.12800744 Report

Quoted By:

Anyone familiar with Pytorch? I'm trying to figure out how to find local max and min of a continuous function

Anonymous

Anonymous Sun 07 Mar 2021 05:56:24 No.12800853 Report

Quoted By: >>12800892

Behold the typical try hard cs major/brianlet
>https://www.youtube.com/watch?v=pKO9UjSeLew
>The starting point could be part of a cycle so he would have to check each element (since he can't mark any) to find feed in's leading to a shitty n^2 algorithm.
>Using a fucking hash table to store the frequency of 0-n
>Sorting instead of doing the blatantly obvious of adapting quick/partition sort to split based on n/2 and see which side is greater in O(n) avg time. To do it in O(1) space just remember the current upper/lower bound and the split size counts for O(nlogn) avg time.
Ugh, watching that was like reading CLRS where they constantly give shitty build up algorithms instead of the optimal one you've already thought of the moment they stated the problem.

Anonymous

Anonymous Sun 07 Mar 2021 06:10:49 No.12800892 Report

Quoted By: >>12800913 >>12801242

>>12800853
Joma is dumb, but it's not bad to build up to the right algorithm with failed attempts. Algorithms are constructive - understanding how one fails and another doesn't is key to making more interesting algorithms, especially when so much of the field comes from combinatorics where knowing the features of certain classes of problems is more important than simply knowing the solution.

Anonymous

Anonymous Sun 07 Mar 2021 06:15:58 No.12800913 Report

Quoted By: >>12801242

>>12800892
>, but it's not bad to build up to the right algorithm with failed attempts
it's an interview strat. always describe the brute force and move up so you don't look completely retarded with nothing to show if you try to build optimal with nothing to stand on

El Arcón

View Same Google iqdb SauceNAO TIMESAND___FractalFire.jpg, 11KiB, 180x252

El Arcón Sun 07 Mar 2021 06:37:27 No.12800965 Report

Quoted By:

>>12800570
This is one of the best images that gets posted on /sci/

Anonymous

Anonymous Sun 07 Mar 2021 08:34:38 No.12801242 Report

Quoted By: >>12802062

>>12800892
>>12800913
Yeah in an interview, you SHOULD explain some dumb straightforward algorithm first before you move on to the optimal one. This also serves as a tactic to buy time.
Unless you are able to get the best algorithm right away.

Anonymous

Anonymous Sun 07 Mar 2021 10:52:52 No.12801565 Report

Quoted By:

anyone done anything with regards to semantic distance? I'm trying to write a python program where you input two keywords, and then the program will scrape the top 100 articles associated with those keywords from a news site (I've written the web-scraping part), and cleaning this text so each keyword has a fat string associated with it. ANd with this data I wanna derive the semantic similarity of the two words; anyone done anything like this or know any particular algorithms or methods for this kind of thing?

Anonymous

Anonymous Sun 07 Mar 2021 14:40:59 No.12802062 Report

Quoted By:

>>12801242
>Unless you are able to get the best algorithm right away.
even if you can, you still just run over brute force and maybe 1 optimization just in case you fuck it up so you don't sound like a confidently wrong retard

Anonymous

View Same Google iqdb SauceNAO 888B1E77-CEC9-44A3-959D-58D29EE4 (...).jpg, 861KiB, 1536x2048

Anonymous Sun 07 Mar 2021 17:26:38 No.12802507 Report

Quoted By: >>12802594 >>12827329

Any logicbros here? I was perusing the CStheory stackexchange and found a post about realizability (can’t post the link because 4chan thinks it’s spam), and I’m thinking of going through the first two books in the first answer:
>Paul Taylor's book Practical Foundations of Mathematics
>Wesley Phoa's notes An Introduction to Fibrations, Topos Theory, the Effective Topos, and Modest Sets
Link for the second thing here (in PS format):
http://www.lfcs.inf.ed.ac.uk/reports/92/ECS-LFCS-92-208/ECS-LFCS-92-208.ps

Has anyone read Paul Taylor’s book? How is it?

Anonymous

Anonymous Sun 07 Mar 2021 17:50:39 No.12802594 Report

Quoted By: >>12802703

>>12802507
The most schizo ramblings you will ever find

Anonymous

Anonymous Sun 07 Mar 2021 18:16:40 No.12802703 Report

Quoted By:

>>12802594
Did you read it? What's so schizo about it?

Anonymous

Anonymous Sun 07 Mar 2021 18:29:00 No.12802759 Report

Quoted By: >>12802832 >>12802998

is the SICP book relevant for me if I went through CLRS and know my way around data structures and algorithms?

Anonymous

Anonymous Sun 07 Mar 2021 18:44:33 No.12802832 Report

Quoted By:

>>12802759
It’s a good book to review and maybe jump into basic compiler / interpreter stuff, but don’t expect to learn a lot of new stuff, especially not algorithmic knowledge.
If you’re feeling rusty on the fundamentals or like you want to brush up, it’s an easy recommendation

Anonymous

Anonymous Sun 07 Mar 2021 19:12:51 No.12802998 Report

Quoted By:

>>12802759
kek no you should probably just skip SICP at that point.

Anonymous

View Same Google iqdb SauceNAO 41YI7ZQd7OL._SX346_BO1,204,203,2 (...).jpg, 32KiB, 348x499

Anonymous Sun 07 Mar 2021 22:36:21 No.12803888 Report

Quoted By:

Anonymous

Anonymous Sun 07 Mar 2021 22:38:05 No.12803896 Report

Quoted By: >>12804757

You guys will probably hate me for this but how do I grind for FAANG jobs?
Is leetcode enough?

Anonymous

Anonymous Mon 08 Mar 2021 01:47:45 No.12804757 Report

Quoted By:

>>12803896
afaik, algorithms is the Great Filter, and leetcode supposedly has problems hard enough to train you for the FAANG interviews. the programming practice is good too because FAANG interviews don't just ask for solutions, they ask for you to implement the solutions in code.

Anonymous

View Same Google iqdb SauceNAO [Shimeji Simulation].jpg, 93KiB, 684x844

Anonymous Mon 08 Mar 2021 04:43:24 No.12805239 Report

Quoted By:

Anonymous

Anonymous Mon 08 Mar 2021 15:46:50 No.12806735 Report

Quoted By:

Anonymous

Anonymous Mon 08 Mar 2021 18:39:53 No.12807502 Report

Quoted By: >>12807613 >>12808026 >>12808556 >>12813106

Why should I learn Computer Science instead of just reading a book on python or C++ and learning by making projects?

Anonymous

Anonymous Mon 08 Mar 2021 18:59:42 No.12807608 Report

Quoted By: >>12808522 >>12808641

>>12796098
>O(max(arr))
yes, which is exponential, because we typically input numbers in a log number of bits.
if we gave number M as M consequent characters, it could be linear, but:
>this is not an actual algorithm on any automaton abstraction I know
>in practice, each computer has an input (large enough) that makes this return false output

Anonymous

Anonymous Mon 08 Mar 2021 19:00:52 No.12807613 Report

Quoted By:

>>12807502
for the millionth time, computer science is not programming.

Anonymous

Anonymous Mon 08 Mar 2021 19:25:01 No.12807725 Report

Quoted By: >>12809081

https://craftinginterpreters.com/

think finishing above is doable in a month?

Anonymous

Anonymous Mon 08 Mar 2021 19:58:38 No.12807839 Report

Quoted By: >>12808522

>>12800192
O(n) integer sorts exist

>>12800347
doesn't it devolve into heapsort then?

Anonymous

Anonymous Mon 08 Mar 2021 20:43:39 No.12808026 Report

Quoted By:

>>12807502
Depends on your goals anon.

Anonymous

Anonymous Mon 08 Mar 2021 23:24:32 No.12808522 Report

Quoted By: >>12808641

>>12800192
>>12800193
>>12807839
have you guys ever even studied any CS? Just look at >>12807608

Anonymous

Anonymous Mon 08 Mar 2021 23:31:52 No.12808556 Report

Quoted By: >>12808739 >>12808744

>>12807502
Why should I learn mathematics instead of just buying a calculator?
That said, pure math is far more worthwhile than CS.
t CS undergrad

Anonymous

Anonymous Mon 08 Mar 2021 23:36:43 No.12808573 Report

Quoted By:

>>12793914
start with learning programming
then, do something like nand2tetris
next... just follow teachyourselfcs.com

Anonymous

Anonymous Mon 08 Mar 2021 23:57:22 No.12808641 Report

Quoted By:

>>12808522
>>12807608
are you saying that O(max(arr)) dominates, which suggests that the size of the input array N drops out? that suggests that no matter what the input size N is you can add more and more to the size and it will be fine
spawning a process is a concurrent operation. if you had N processors then yes this is O(N + max(arr)) and those two terms are independent of each other

if you didn't have unlimited resources as N grows, you put more load on your OS's process manager which has to execute all of those spawned N processes. you're bound by how well your processor could actually execute N processes
degenerate case is if you had an array all consisting of the number 1. As N grows you spawn more processes and eventually you will take more than 1 second to execute unless you keep adding CPUs

Anonymous

Anonymous Tue 09 Mar 2021 00:30:39 No.12808739 Report

Quoted By: >>12808744

>>12808556
>That said, pure math is far more worthwhile than CS.
Nah, especially not when there's a lot of pure math that was given reinvigorated motivation after CS. We would have expanders without theoretical computer science, which means in turn we wouldn't have nearly as rich a theory of thin groups, which means pure number theory will have lost an incredibly interesting topic. CS both utilizes and contributes to pure math as a whole.

Anonymous

Anonymous Tue 09 Mar 2021 00:33:12 No.12808744 Report

Quoted By: >>12810217

>>12808556
>>12808739
not saying pure math isn't worthwhile though. Just saying it's unfair to throw CS under the bus. The activity in CS and physics in quantum information these days has led a lot of attention to operator theory as well.

Anonymous

Anonymous Tue 09 Mar 2021 02:24:19 No.12809081 Report

Quoted By: >>12809115

>>12807725
Yeah it’s a guide. Take what you need from it and make your own thing. The code is all there for you, you could “do” it in a day.

Anonymous

Anonymous Tue 09 Mar 2021 02:33:34 No.12809115 Report

Quoted By: >>12809963

>>12809081
i want to do it in python just because i'm too lazy to set up the entire jvm
i just wanna run through it once and then work on my own shit after

Anonymous

Anonymous Tue 09 Mar 2021 07:11:34 No.12809963 Report

Quoted By: >>12809965 >>12819296

>>12809115
It’s very OOP/Java centric, but you can pick up the basics. If you’re interested in compilers/interpreters, I’d just learn a little theory. Lexers are just finite state automata and LL(1) parsers like the one I think that book uses are very simple to think about once you understand context-free grammars and PDA’s. That being said, assess your goals. If you want an easy parser to read, try Luaparse. It’s in JavaScript but it’s terse and easy to follow.

Anonymous

Anonymous Tue 09 Mar 2021 07:13:04 No.12809965 Report

Quoted By:

>>12809963
Moreover, repl.it has a Java mode. You could always use that.

Anonymous

Anonymous Tue 09 Mar 2021 09:03:00 No.12810217 Report

Quoted By: >>12810407

>>12808744
From my experience, math classes teach you far more than CS ones. Not because the major is intentionally half-assed, but because of what kind of people go into CS. I mean, just look at this thread, it's a mess how these guys reason about things.
A masters in analysis will make you far more based than whatever major in CS.

Anonymous

Anonymous Tue 09 Mar 2021 10:31:42 No.12810407 Report

Quoted By: >>12810513

>>12810217
>From my experience, math classes teach you far more than CS ones.
I mean at the undergrad level sure, but your point:
>A masters in analysis will make you far more based than whatever major in CS.
1) nobody hands out a masters (MSc) in analysis lmao
2) there's no reason to believe a graduate degree in theory focused CS involves any less mathematical maturity than a graduate degree focused on discrete mathematics, and even so, there's a ridiculous amount of nondiscrete math in TCS these days anyway.

So I agree that math helps a lot and even subsumes CS at the undergrad level, in grad it's not so clear cut
>t. double majored in math/CS, in math grad now

Anonymous

Anonymous Tue 09 Mar 2021 11:17:25 No.12810513 Report

Quoted By: >>12820583 >>12820602

>>12810407
>nobody hands out a masters (MSc) in analysis lmao
maybe not in the US, but in Europe, it's a different story. Most people actually go for a masters and don't do PhD.

Analysis has many advantages to CS:
>it is far older and far more developed
>high quality study materials are out there and there's never a barrier in smooth learning other than not having the intellectual capacity
>the knowledge culminates in a sort of a layered way, every new thing you learn builds on much of what you've covered before, so that your final knowledge is very 'deep', whereas much of CS or even discrete math has to be learned ad hoc, so the knowledge isn't 'deep'
>it's [spoiler:lit]way more fun[/spoiler:lit] partly because of the previous points

Anonymous

View Same Google iqdb SauceNAO 1654156467435.jpg, 80KiB, 500x500

Anonymous Tue 09 Mar 2021 11:20:45 No.12810523 Report

Quoted By: >>12812411

>>12793853
>Math for CS
Is there even such a thing? It can require knowledge on any field depending on the goal, no? I guess these books are more like a primer?

?????? !!Tfq3QaHoS4a

View Same Google iqdb SauceNAO Temporary workaround.jpg, 493KiB, 1561x2048

?????? !!Tfq3QaHoS4a Tue 09 Mar 2021 14:37:14 No.12811030 Report

Quoted By: >>12812106 >>12812243 >>12812394 >>12813628 >>12813707 >>12821776 >>12823391 >>12827239 >>12831904

>>12793853
Due to real life obligations and time constraints, change in job industry (??? -> finance), and unfortunate health concerns (*coff*), I've decided to shut down the server. I might set it up again in a few months, but I wouldn't hold your breath. Sorry for wasting everyone's time

Anonymous

View Same Google iqdb SauceNAO 1614520032583.png, 398KiB, 448x478

Anonymous Tue 09 Mar 2021 19:52:55 No.12812106 Report

Quoted By:

>>12811030
sucks that it's gone, but you gotta do what you gotta do i guess

Anonymous

View Same Google iqdb SauceNAO oYYpuoWQmAQ.jpg, 43KiB, 550x405

Anonymous Tue 09 Mar 2021 20:25:53 No.12812243 Report

Quoted By:

>>12811030
Anon :(
Did you really have to do this?

Anonymous

Anonymous Tue 09 Mar 2021 21:03:29 No.12812394 Report

Quoted By:

>>12811030
I was just gonna stop by to rant and ask why it's gone. Why not just transfer ownership to someone trustworthy?

Anonymous

View Same Google iqdb SauceNAO 5.jpg, 79KiB, 720x960

Anonymous Tue 09 Mar 2021 21:08:16 No.12812411 Report

Quoted By: >>12812499 >>12812963

>>12810523
Math for CS is basically just discrete mathematics. You can get away with just understanding base 2 logarithmic functions and exponents. Even then, it's just used for algorithmic analysis, where you're just testing scalability. CS Math isn't hard at all but it certainly helps if you're working on anything super advanced.

Anonymous

Anonymous Tue 09 Mar 2021 21:32:53 No.12812499 Report

Quoted By:

>>12812411
>what is the whole of optimization
>what are probabilistic algorithms?
>what is signal processing?
>what is digital image processing?
Undergrads shouldn't feel entitled to talk about what they don't understand

Anonymous

Anonymous Tue 09 Mar 2021 21:59:47 No.12812604 Report

Quoted By: >>12816974

>>12793853
The Discord server disappeared?

Anonymous

Anonymous Tue 09 Mar 2021 23:44:56 No.12812963 Report

Quoted By:

>>12812411
>it's just discrete mathematics
>Even then, it's just used for algorithmic analysis, where you're just testing scalability.
>CS Math isn't hard at all but it certainly helps if you're working on anything super advanced.
Anon you don't really
most CS programs suck, but this is massive dunning kruger

Anonymous

Anonymous Tue 09 Mar 2021 23:58:57 No.12813023 Report

Quoted By: >>12813251

>>12793853
Looking to learn Digital Image Processing and Computer Graphics. Which topic should I start with and with books should I use? Shirley or Hughes et al? Also does anyone have any historical and intuitively motivated color science/colorimetry recommendations?

Anonymous

Anonymous Wed 10 Mar 2021 00:13:08 No.12813076 Report

Quoted By: >>12820638

>>12794163
>Driving isn't considered a real field by /sci/ standard.
>Too little hard engineering for eng, too little math for math, and too little real lab work for science.
>The funny thing is half of the math people probably need to learn how to Drive to make money anyway.

Anonymous

Anonymous Wed 10 Mar 2021 00:16:06 No.12813085 Report

Quoted By: >>12813380

>>12796466
>How delusional are you?
Just because you define shit with sets and use induction doesn't make it hard core math.

Anonymous

View Same Google iqdb SauceNAO CS guide.png, 178KiB, 1068x1142

Anonymous Wed 10 Mar 2021 00:22:22 No.12813106 Report

Quoted By: >>12813236

>>12807502
>Why should I learn Computer Science
In university? You shouldn't. Just study it on your own.

Anonymous

Anonymous Wed 10 Mar 2021 00:51:55 No.12813236 Report

Quoted By: >>12813279 >>12816443

>>12813106
The steps are easy but takes like 1 year for each

Anonymous

Anonymous Wed 10 Mar 2021 00:55:52 No.12813251 Report

Quoted By:

>>12813023
> Which topic should I start with and with books should I use?
For sigproc, I liked using Berkeley and MIT's notes from their EECS programs / teacher webpages. For graphics, CMU has wonderful resources. Keenan Crane's differential geometry course is really good.
http://graphics.cs.cmu.edu/?page_id=16

Anonymous

Anonymous Wed 10 Mar 2021 01:01:05 No.12813279 Report

Quoted By:

>>12813236
Only if you're a brainlet.

Anonymous

Anonymous Wed 10 Mar 2021 01:18:09 No.12813367 Report

Quoted By:

>>12793914
FORTRAN

Anonymous

View Same Google iqdb SauceNAO CMU_DDG_FA17_A1_Coding_icon.jpg, 70KiB, 1000x779

Anonymous Wed 10 Mar 2021 01:21:04 No.12813380 Report

Quoted By: >>12816344 >>12821066

>>12813085
>hard core math
>engineering
the fuck are you smoking? "Advanced engineering mathematics" is the baby calculus used to solve a basic heat equation.
At least at a more than decent CS program, you do something like Keenan Crane's Differential Geometry (https://brickisland.net/DDGSpring2021/) for graphics and cool shit. Most CS programs are shit, but at least the good ones give a much more proper treatment of the math. Pic related

Anonymous

Anonymous Wed 10 Mar 2021 01:24:47 No.12813396 Report

Quoted By: >>12813628

>>12793853
>Discord: https://discord.gg/S2j9jJJGKM
Says it is invalid, I'd like to join the Concrete Mathematics study group.

Anonymous

Anonymous Wed 10 Mar 2021 02:28:55 No.12813628 Report

Quoted By: >>12813688

>>12813396
See >>12811030

Anonymous

Anonymous Wed 10 Mar 2021 02:50:12 No.12813688 Report

Quoted By: >>12813710

>>12813628
That sucks, I wanted to join one of these groups for the first time :(
Surely there must be other discord groups for this, right?

Anonymous

Anonymous Wed 10 Mar 2021 02:58:01 No.12813707 Report

Quoted By:

>>12811030
Hey dude, take care of yourself. Thanks for making this for people.
Keep people in the loop if you can. All the best my dude

Anonymous

Anonymous Wed 10 Mar 2021 02:58:39 No.12813710 Report

Quoted By:

>>12813688
Check out /bpg/ on /g/, threads don't get posted often but the Discord still is somewhat active, albeit slow

Anonymous

Anonymous Wed 10 Mar 2021 06:47:43 No.12814360 Report

Quoted By: >>12814562

What's a supplement book for sipser?

Anonymous

Anonymous Wed 10 Mar 2021 07:54:26 No.12814521 Report

Quoted By: >>12814701

What is better Sipser or Hopcroft + Ullman?

Anonymous

Anonymous Wed 10 Mar 2021 08:10:46 No.12814562 Report

Quoted By:

>>12814360
Kozen

Anonymous

Anonymous Wed 10 Mar 2021 09:40:45 No.12814701 Report

Quoted By: >>12817149

>>12814521
Sakarovitch for a non-brainlet intro.

Anonymous

Anonymous Wed 10 Mar 2021 20:20:44 No.12816344 Report

Quoted By:

>>12813380
Why the duck does everyone ignore graphics in these threads. Easily one of the coolest and most immediately appreciable applications of CS.

Anonymous

Anonymous Wed 10 Mar 2021 20:41:43 No.12816443 Report

Quoted By: >>12817114

>>12813236
That's the joke, each step is basically a full course

Anonymous

Anonymous Wed 10 Mar 2021 22:47:59 No.12816915 Report

Quoted By:

bros...

Anonymous

Anonymous Wed 10 Mar 2021 23:10:19 No.12816974 Report

Quoted By: >>12823720

>>12812604
Yes, the admin is apparently a fucking retard.

Anonymous

Anonymous Wed 10 Mar 2021 23:31:22 No.12817049 Report

Quoted By:

how are these study groups being organized? I'd be down to join SICP, PAPL or Concrete Maths

Anonymous

Anonymous Wed 10 Mar 2021 23:56:40 No.12817114 Report

Quoted By:

>>12816443
You wouldn’t believe the amount of people who share it around like it’s something you ought to do though

Anonymous

Anonymous Thu 11 Mar 2021 00:09:27 No.12817129 Report

Quoted By: >>12820118

>>12798026
there's math in programming language theory? like what? at the introductory undergrad level i mean obv.

I was thinking about double majoring but most of the upper-div required classes looked pretty gay at my school. they are
>computer ethics (gay)
>a couple capstones (bootcamp)
>'elements of software-eng', (bootcamp)
>princiciples of programming languages (syllabus reads more like 'intro to paradigms other than OO')
>algo analysis/complexity (worthwhile)
>operating systems (worthwhile)
>intro to databases (bootcamp)
>EE hardware class (they use the NAND2TETRIS book, actually looks good)

it added a whole year for what I thought was majority bootcamp-type stuff. am I wrong here? the only classes that I thought mattered were classes I plan on continuing to self teach (copying a variant of Ben Eater's 8-bit computer build instead of NAND2TETRIS, and will read some OS book later). and the electives offered were mostly super dull/boring as well. I just struggled to piece together a cs major that I was as excited about as I am my math stuff. what are you taking for math, by the way? I've had a hard time picking the right mix between pure and employable.

so I'm not sure yet, so im minoring for now and just taking alog/complexity along with numerical comp, machine learning (I,II), and rando algos & probabilistic analysis. figure I'll add the year at the end if I really think it's worth it. my math classes on the other hand, couldn't be more excited about: Real/Complex Anal, Abstract Alg, Topology, Diff Geo, and a couple more applied classes like numerical optimization and stats just to round things off. I'm hoping they let me take a computational geometry class, but it's only offered graduate level.

Anonymous

Anonymous Thu 11 Mar 2021 00:10:37 No.12817132 Report

Quoted By: >>12818058 >>12819009

>>12798030
the SICP authors are dropping a new book this month that's supposed to be about this kinda stuff

https://mitpress.mit.edu/books/software-design-flexibility

looks like a good read

Anonymous

Anonymous Thu 11 Mar 2021 00:16:29 No.12817149 Report

Quoted By: >>12817612

>>12814701
i'm a brainlet

Anonymous

Anonymous Thu 11 Mar 2021 03:11:19 No.12817612 Report

Quoted By:

>>12817149
Sipser is fine

Anonymous

Anonymous Thu 11 Mar 2021 05:53:10 No.12818058 Report

Quoted By:

>>12817132
Oh shit, that’s really cool

Anonymous

Anonymous Thu 11 Mar 2021 13:10:59 No.12819009 Report

Quoted By:

>>12817132
>https://mitpress.mit.edu/books/software-design-flexibility

You can already download it from z lib

Anonymous

Anonymous Thu 11 Mar 2021 15:07:50 No.12819296 Report

Quoted By:

>>12809963
mainly i want to know enough so i can get to IR optimization

Anonymous

Anonymous Thu 11 Mar 2021 19:07:51 No.12820097 Report

Quoted By:

>>12793853
Anyone have any resources for hard dynamic programming questions or learning novel uses of dynamic programming?

Anonymous

Anonymous Thu 11 Mar 2021 19:12:29 No.12820118 Report

Quoted By:

>>12817129
No one cares bro

Anonymous

Anonymous Thu 11 Mar 2021 19:18:07 No.12820138 Report

Quoted By:

>>12796840
>ML will just keep moving forward.
>that pic
kek

Anonymous

Anonymous Thu 11 Mar 2021 21:12:42 No.12820583 Report

Quoted By:

>>12810513
>it is far older and far more developed
Analysis is very mature, yes, but that's kind of the point - you study CS because you aim to mature it because it's only around 60 years old as far as the academic tradition goes. Even then, it's started maturing rapidly. This is the point of research.
>so that your final knowledge is very 'deep', whereas much of CS or even discrete math has to be learned ad hoc, so the knowledge isn't 'deep'
This is patently false though. There is a very well studied bed of mathematical logic, combinatorial theory (don't give me the "it isn't structured" moniker - there's a clear combinatorial theory that extends from basic counting to the theory of hypergraphs and design, etc etc.) that underlies the foundation of computer science.
Hell, traditional analysis shows up in nontrivial ways in concentration of measure used in algorithms these days anyway.

Anon, it sounds like you're more biased towards a field where having a "story" or canonical presentation and pedagogy is more valuable than the actual theory or what it's studying. I *like* analysis a lot, but it feels like you're focusing on the aesthetics of how it feels to be a student rather than the reality of being a research, whether it's in analysis or some part of computer science.

Anonymous

Anonymous Thu 11 Mar 2021 21:18:08 No.12820602 Report

Quoted By:

>>12810513
>I like analysis because it has more prestige attached and it seems fancier: the post
that feeling quickly goes away when you start actually doing analysis professionally. you better actually like it rather than jerking yourself off
t. analysisfag
>the knowledge isn't deep
HoTT begs to differ

Anonymous

View Same Google iqdb SauceNAO 1585995834587.jpg, 27KiB, 386x402

Anonymous Thu 11 Mar 2021 21:30:19 No.12820638 Report

Quoted By:

>>12813076
average CS basher

Anonymous

Anonymous Thu 11 Mar 2021 23:30:27 No.12821051 Report

Quoted By:

>>12793879
>automated regression is a bubble
>descriptive statistics is a bubble

You might as well say "science" is a bubble.

Anonymous

Anonymous Thu 11 Mar 2021 23:31:41 No.12821057 Report

Quoted By: >>12821438

>>12796107

Make yourself richer. Build your own cluster for computing services.

Anonymous

Anonymous Thu 11 Mar 2021 23:33:27 No.12821066 Report

Quoted By: >>12821736

>>12813380

What is the actual purpose of the diagram? What do we gain from diagrammatic communication? Why couldn't whatever the diagram is saying be said with words like, "this composition of functions is analogous to this composition of functions"

Anonymous

Anonymous Fri 12 Mar 2021 01:37:24 No.12821438 Report

Quoted By: >>12825341

>>12821057
>Make yourself richer.
if you just wanted money you just become an SWE and make iphone apps for 3x the income as a researcher

Anonymous

Anonymous Fri 12 Mar 2021 03:44:27 No.12821720 Report

Quoted By: >>12821742

>kicked out the discord
k didn't know i had to be active everyday

Anonymous

Anonymous Fri 12 Mar 2021 03:49:03 No.12821736 Report

Quoted By:

>>12821066
>"this composition of functions is analogous to this composition of functions"
It has a real use here, since it's showing how to characterize some salient differential geometry invariant of whether it shows up in the continuous or the discrete cases. Read up on why people use commutative diagrams in the first place. You put it in this form to lay plain the algebraic structure without *really* worrying too much about these objects themselves, just that you can compose them with maps with certain properties. See things like the Five Lemma to see applications.

Read the associated lecture here: https://brickisland.net/DDGSpring2019/wp-content/uploads/2019/02/DDG_458_SP19_Lecture09_DiscreteExteriorCalculus.pdf
The commutative diagram is used to express that we can talk about grad, curl, div, etc. using compositions of these operators on the k-forms / their duals.

Anonymous

Anonymous Fri 12 Mar 2021 03:50:04 No.12821742 Report

Quoted By: >>12821745

>>12821720
the discord was shut down. you weren't kicked

Anonymous

Anonymous Fri 12 Mar 2021 03:52:00 No.12821745 Report

Quoted By: >>12821776

>>12821742
oh why was it shut down?

Anonymous

Anonymous Fri 12 Mar 2021 04:08:15 No.12821776 Report

Quoted By: >>12821789

>>12821745
see >>12811030

Anonymous

Anonymous Fri 12 Mar 2021 04:13:58 No.12821789 Report

Quoted By: >>12824799

>>12821776
rip

Anonymous

Anonymous Fri 12 Mar 2021 14:37:12 No.12823046 Report

Quoted By:

>>12794163
/sci/ is filled with shit for brains shizos and undergrads, who cares about what /sci/ thinks.

Anonymous

Anonymous Fri 12 Mar 2021 16:50:32 No.12823367 Report

Quoted By:

>>12793879
>Anyone think the ML bubble is gong to pop?
Yes, even Minsky himself said Neural Nets are fatally flawed will hit a hard brickwall in the future. They physically cannot look at everything that they know, that is, they can only look at their knowledge base through a sort of "tunnel vision"; obviously problematic for scalability.
We are already hitting a point of diminishing returns, take a look at pic related, look at how many parameters GPT-3 has compared, and yet GPT-3 suffers from the same fatal flaws as GPT-2 (can't comprehend for shit).
The money that is flowing into Machine Learning is expecting the technology to keep improving, what do you think will happen once when companies realize that their AI investments aren't doing as good as they used to? Add all the hype and high expectations people have for the technology and we've nearly checked every box for an AI winter.

Anonymous

View Same Google iqdb SauceNAO 1614442329283.png, 15KiB, 760x682

Anonymous Fri 12 Mar 2021 16:51:35 No.12823371 Report

Quoted By:

>>12793879
>Anyone think the ML bubble is gong to pop?
Yes, even Minsky himself said Neural Nets are fatally flawed will hit a hard brickwall in the future. They physically cannot look at everything that they know, that is, they can only look at their knowledge base through a sort of "tunnel vision"; obviously problematic for scalability.
We are already hitting a point of diminishing returns, take a look at pic related (sorry it's a bit blurry), look at how many parameters GPT-3 has compared, and yet GPT-3 suffers from the same fatal flaws as GPT-2 (can't comprehend for shit).
The money that is flowing into Machine Learning is expecting the technology to keep improving, what do you think will happen once when companies realize that their AI investments aren't doing as good as they used to? Add all the hype and high expectations people have for the technology and we've nearly checked every box for an AI winter.

Anonymous

View Same Google iqdb SauceNAO 1587841285555.png, 15KiB, 760x682

Anonymous Fri 12 Mar 2021 16:52:40 No.12823377 Report

Quoted By: >>12823760

>>12793879
>Anyone think the ML bubble is gong to pop?
Yes, even Minsky himself said Neural Nets are fatally flawed will hit a hard brickwall in the future. They physically cannot look at everything that they know, that is, they can only look at their knowledge base through a sort of "tunnel vision"; obviously problematic for scalability.
We are already hitting a point of diminishing returns, take a look at pic related (sorry it's a bit blurry), look at how many more parameters GPT-3 has, and yet GPT-3 suffers from the same fatal flaws as GPT-2 (can't comprehend for shit).
The money that is flowing into Machine Learning is expecting the technology to keep improving, what do you think will happen once when companies realize that their AI investments aren't doing as good as they used to? Add all the hype and high expectations people have for the technology and we've nearly checked every box for an AI winter.

Anonymous

Anonymous Fri 12 Mar 2021 16:57:24 No.12823391 Report

Quoted By:

>>12811030
good, next time try IRC instead faggot.

Anonymous

Anonymous Fri 12 Mar 2021 18:19:04 No.12823720 Report

Quoted By:

>>12816974
Always was.

Anonymous

Anonymous Fri 12 Mar 2021 18:26:47 No.12823760 Report

Quoted By:

>>12823377
the hard problem with neural nets is finding the right graph topology + weights for n billion nodes, imo this can only be brute forced through an evolutionary algo which would require at least 10+ years of hardware improvements to be achievable

Anonymous

Anonymous Fri 12 Mar 2021 23:25:06 No.12824799 Report

Quoted By:

>>12821789
Yeah

Anonymous

Anonymous Sat 13 Mar 2021 02:15:24 No.12825341 Report

Quoted By: >>12826117

>>12821438
not exactly true. Doing research in a hot field is still big bucks and high impact.
becoming an SWE can sometimes be a path to big bucks without working too hard, but it's relentlessly boring work most of the time imo

Anonymous

Anonymous Sat 13 Mar 2021 06:49:47 No.12826117 Report

Quoted By: >>12828024

>>12825341
research probably pays well but i can't imagine going into it with money in mind
i've noticed a trend of csfags going for a ms/phd later since they don't want to pass up the money in SWE

i'd do that myself but i also graduated at 24 so i don't have as much time to fuck around as most SWEs. either i get into research soon or i never get in

Anonymous

Anonymous Sat 13 Mar 2021 08:50:54 No.12826313 Report

Quoted By:

I plan to actually buy some books whenever I get my stimulus money. I have been reading PDFs and .djvu files etc, which is better than nothing, but I really prefer a book to staring at a screen. For one, it’s easier to focus. Also, the quality of scans varies greatly. It ranges from "good enough" to "practically unreadable". In particular, I notice characters (,),|,",' are often distorted or outright missed. This isn’t always a problem as you can usually infer what was meant to be there, but in the lower quality scans, or when working with unfamiliar material, it is a huge hinderance. It’s really painful when the typeset for an equation is difficult to interpret. There is also a certain satisfaction with owning the actual book.

Anonymous

Anonymous Sat 13 Mar 2021 16:39:43 No.12827239 Report

Quoted By: >>12827319

>>12811030
/g/'s /dpt/ server is taking refugees
/4TuhPhr
I'll probably regret posting it on /sci/ but hell let's give it a try

Anonymous

Anonymous Sat 13 Mar 2021 17:06:37 No.12827319 Report

Quoted By:

>>12827239
Might be a good place for people getting into programming, but not necessarily the best place for people learning CS.

Anonymous

Anonymous Sat 13 Mar 2021 17:08:08 No.12827329 Report

Quoted By: >>12827453

>>12802507
Looks to be a category theory treatment of mathematical logic which is fine I guess though not standard, the standard book for this kind of thing is Kleene's or Shoenfield's math logic books you can get on library genesis, and many other authors with similar books. Those kinds of books typically cover the idea of recursion, godel's theorems, model theory which is pretty interesting. First order logic and set theory are not interesting at all however

The HoTT book if you read it covers some foundations, then you would just go into modern computational type theory to see what they are doing there lately if you wanted to mix that category theory book with recent developments.

Anonymous

Anonymous Sat 13 Mar 2021 17:50:57 No.12827453 Report

Quoted By: >>12830078 >>12830094

>>12827329
I’ve only read 2 chapters of Shoenfield and a couple more of Enderton. Do you think I should go further with Taylor’s book or just finish more of Shoenfield first, especially since algebraic topology (up to cohomology, particularly Poincare duality) is my only experience with category theory?
>HoTT
oh right of course. I’ve read up to dependent types. Is there literature (particularly papers) you’d recommend?

Anonymous

Anonymous Sat 13 Mar 2021 20:32:09 No.12828024 Report

Quoted By: >>12828786

>>12826117
why do you say this? is there some sort of age cap for doing research or going for a phd? im currently in my late 20s about to finish my undergrad/masters (started late) and have been doing research for the most part but doing a SWE job for shits and giggles right now

Anonymous

Anonymous Sat 13 Mar 2021 20:50:38 No.12828088 Report

Quoted By:

>>12793853
Discord invite has expired

Anonymous

Anonymous Sat 13 Mar 2021 23:28:59 No.12828786 Report

Quoted By:

>>12828024
>why do you say this? is there some sort of age cap for doing research or going for a phd?
from an objective standpoint, no. it's not uncommon and there's lots of people that go to grad school in their late 20s or early 30s. it's arguably the better choice because you go in with more money and less weight on your shoulders about just "finding a job"
if you're capable of doing a phd, you could have become a staff or principal making 400k+ instead of doing research. they're doing it because they want to do it, not because it pays more

i graduated kind of late so i have less experience overall. this field skews young, and being old fucks with your mental state pretty hard. most people my age who follow my path make triple my income, i am 4 years behind the curve.

> im currently in my late 20s about to finish my undergrad/masters (started late) and have been doing research for the most part but doing a SWE job for shits and giggles right now
are you working as an SWE while in school or something?

Anonymous

Anonymous Sun 14 Mar 2021 05:59:11 No.12830078 Report

Quoted By: >>12830094 >>12830222

>>12827453
Google the names in this talk and read whatever they are doing presently: http://www.cs.cmu.edu/~rwh/talks/MURI20-talk.pdf basically they dumped HoTT for being insufficiently computational, came up with 'cartesian computational higher type theory' and have alpha software on github to write a proof assistant already.

There's lectures on this in the OPLSS archive like this one https://www.cs.uoregon.edu/research/summerschool/summer18/lectures/harper1-1.mp4 he has a draft book you could prob ask for and get

Anonymous

Anonymous Sun 14 Mar 2021 06:03:37 No.12830094 Report

Quoted By: >>12830222

>>12830078
>>12827453
Oops, btw everything he is writing on the board is contained here for background: http://symbolaris.com/course/constlog-schedule.html

Anonymous

Anonymous Sun 14 Mar 2021 06:42:22 No.12830222 Report

Quoted By: >>12832359

>>12830078
>>12830094
Thanks a lot for the resources! I skimmed through the presentation - did they “dump” HoTT entirely or did they just take issue with the computational meaning / realizability?

Anonymous

Anonymous Sun 14 Mar 2021 16:01:24 No.12831756 Report

Quoted By: >>12831904

Is the discord server still up? Can anyone post an invite?

Anonymous

Anonymous Sun 14 Mar 2021 16:41:30 No.12831904 Report

Quoted By:

>>12831756
Anon, I...
>>12811030

Anonymous

Anonymous Sun 14 Mar 2021 16:53:58 No.12831950 Report

Quoted By: >>12833121 >>12833840

Anyone wanna try starting another server?

Anonymous

Anonymous Sun 14 Mar 2021 18:20:04 No.12832359 Report

Quoted By: >>12832612

>>12830222
Really if you want to learn what they are doing is to read everything Voedvodsky wrote about univalent foundations https://www.ias.edu/ideas/2014/voevodsky-origins which was to replace category theory and ZFC with some new kind of foundations for math he could just run in a computer. This idea takes off then a bunch of type theorists come up with HoTT, which is an axiomatic formalism ergo not sufficiently computational so they came up with many different cubical models of HoTT and their own cubicalTT that is fully explained here: https://www.cs.cmu.edu/~cangiuli/papers/abcfhl.pdf

Many of these extend or model HoTT so you still have to know the HoTT book and ideas: https://scs.hosted.panopto.com/Panopto/Pages/Sessions/List.aspx#folderID=%2207756bb0-b872-4a4a-95b1-b77ad206dab3%22

The current state of univalence by Voedvodsky is there is now a bunch of different cubical models, a bunch of different proof assistants, and of course they are using some kind of model theory I know nothing about to prove all the models are the same, and to prove the type checking algorithms which I believe they already did, but I haven't kept up with whatever is going on. This all sounds highly technical but the proof assistants are pretty easy to use, the cubical models themselves simple to understand if you know some basic category theory. You don't have to know anything about type theory except when they use some jargon like 'product types' then you look in this book to see what it is: http://www.cs.cmu.edu/~rwh/pfpl/index.html

Anonymous

Anonymous Sun 14 Mar 2021 19:03:57 No.12832612 Report

Quoted By:

>>12832359
I see, that makes sense at a high level view. Would you say that there are more people in the math departments working on this or more people in the CS departments? I’m not biased towards either since I majored in both, but I’m wondering what department I might want to apply to when I apply for grad

Anonymous

Anonymous Sun 14 Mar 2021 20:58:04 No.12833121 Report

Quoted By:

>>12831950
Dunno, but I could try continuing on the CS threads

Anonymous

Anonymous Sun 14 Mar 2021 23:38:05 No.12833840 Report

Quoted By:

>>12831950
what are these servers like anyways?

Anonymous

Anonymous Mon 15 Mar 2021 00:14:01 No.12833953 Report

Quoted By: >>12834607 >>12834950

>>12800570
man, it looks sexy. too bad i'm too dumb for it

Anonymous

Anonymous Mon 15 Mar 2021 03:32:55 No.12834607 Report

Quoted By:

>>12833953
You could learn it if you wanted to

Anonymous

Anonymous Mon 15 Mar 2021 05:37:49 No.12834950 Report

Quoted By:

>>12833953
just do it

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