Machine Learning/AI

>>61302492
Big data & ML/neural networks, yep.

>>61302492
It's a stupid cuck meme. Practically no one needs to learn ML because no one has an actual application for it. You can just use APIs.

Learn Java enterprise dev if you actually want a paycheck.

>>61302492
Real question is should i use tensorflow or something on top of hadoop? i think big data is the most important trend to optimize for

>>61302518
>Java
>2000+17

ML is going to replace a lot of computer science.
The reason you shouldn't get into it is that it's already becoming more of a service and less of a development problem.

Why would someone hire you to build an image classifier when they can just finetune the last layer of a pretrained resnet and obtain more accuracy than you would be able to?

>>61303340
>ML is going to replace a lot of computer science.
What for example?

>>61302518
Get out of /g/

>>61304846
one example of a field i'm interested in is malware research. AVs need to die, they just aren't effective in the current environment. As soon as someone figures how to apply RNNs to binaries in realtime, that's when AVs will begin to die.

File: anon.jpg (14KB, 731x113px) Image search: [Google]

14KB, 731x113px

>>61302492
>ML
>Computer science
Nope, my dude, actual ML research is based on statistics, convex analysis and game theory.
The engineering is mostly a bottleneck in ML and not much more, the whole field is in general worngly associated to CS because >muh computers and >muh kode.

I took ML up during my Master's in CS but I had
to pivot HARD to actually be able to work in research. Again, universities tend to group ML and AI along with other CS disciplines, but it really is more affine to pure mathematics, statistics or neuroscience.

Refer to >>61303340 for what a "computer scientist" can do with ML, as you'll mostly be using an API.

tl;dr don't pick it up if you don't know what a vector space and a gradient are.

>>61304909
>Cs babies literally so stupid they can't understand 1st year uni math courses

And you wonder how pajeets that value math make it farther

>>61302492
"AI Engineer" is the new buzzword, ML is just a part of AI and only took off because finally we have large enough datasets to do something with them.

Norvig's AI book is still the goto reference for such things and the free deeplearning book (google it). His old AI book written in Lisp is one of the best programming books ever made in terms of leveling up to a more advanced programmer.

If you're interested in ML specifically and you have a basic programming background plus basic linear algebra then here you go: https://functionalcs.github.io/curriculum/#org0cfab06

>>61302518
It's true that you just use APIs, but you still need a degree to get hired in the first place. Also, it helps to know which method to use under given conditions.

>>61305052
Where did I say I don't understand it, little faggot?
I just said the degree isn't centered on that so you have to integrate.

>>61302518
>bait

Wallstreet has hired "computational statistics and probability" which is essentially what ML is, since the early 1980s.

Anybody who has a degree is stats or a compsci grad with just one or two courses in ML/stats will be in high demand on Wallstreet.

>>61305314
>The cs degree isn't centered on it
Like I said, cs majors literally so stupid that they do not understand the math every other stem (minus the m) degree needs to take year 1

File: 1472786063002_0.png (14KB, 500x500px) Image search: [Google]

14KB, 500x500px

>>61305408
>Isn't centered
>It's not covered
Learn the difference faggot, at my uni we take the same exact courses as everyone else in the first two years.
Jesus Christ how insecure can you be about your competence? Scared you'll be out of a job soon, once stupid CS dudes will replace your sorry ass with universal approximators?

>tfw wanting to build a deep learning workstation but no GPUs in stock
Fuck miners.

>>61305487
>Cs baby that barely passed calc 2 and linear algebra replacing higher level math

Fucking rolf

>>61305554
>/sci/ virgin thinks that his group theory or whatever circle jerk is somehow relevant
Got news for you pal: math and physics research has been dead for the past 40 years. Should I throw you some money so you can dry them tears off?

>>61306520
>What is modern numerical analysis, stochastic modeling and predictions in physics and finance, all mixed with statistics

Cs brainlets thinking they're going to make it into machine learning and AI are the equivalent of highschool gamer-tards wanting to go into cs to make games.

Cs undergrades are so delusional in their intellect and how hard cs is and I fucking lol when they can't pass linear algebra.

Any good math program deals with "real world applications" and study useful algorithms much more in-depth and involved than cs idiots bragging about that they know how binary search works

>>61306937
Shit how could I forget optimization (linear, nonlinear, discrete), probably the most relevant to cs

File: 1488420514820.jpg (116KB, 1200x948px) Image search: [Google]

116KB, 1200x948px

>>61306937
>I s-swear guise, i didn't waste my life sucking cock in academia
Show me stats, fagboi, I can't hear you over all the money society is throwing at me for being more useful than you.

Also
>Stochastic modeling and predictions in physics and finance
Lol you basically described one application machine learning without even knowing it. By the time you have pulled some weird abstraction out of your ass I have fit a billion different MLP 100x better than your shitty engineered prediction model.

How can /sci/tards even compete? Go back to the IQ threads, you little shit, you're done here. Even the worst scumbag pedo freetard of /g/ is better than you.

Posting rare Terrys to humiliate this faggot moar

>>61307302
Ok listen here brainlet, the point I'm making is that ML is more closely linked to math than cs.
Good math programs focus on scientific computing and many cs programs don't need more math than linear algebra.

So tell me, how is a cs grad that only took linear algebra going to implement all these high level math applications in ML and refine them?

>>61307395
Well fagtron, I don't know how, but that's exactly what I am doing as main focus of my master and will continue to do for my PhD.
You on the other hand... What have you contributed to society with your useless circlejerk? Did the Prof let you suck his dick without a condom?

File: 1493144112997.jpg (37KB, 540x540px) Image search: [Google]

37KB, 540x540px

>>61307534
>I dont know how but that is my main focus
>The focus of my masters is on undergrad math classes
>But math is useless to machine learning and to society

>>61307692
Fampai, your logic is starting to fade.
Math is useful for ML, and ML is useful to society.
Pure math degrees on the other hand... O boi am I LAFFIN @ u

File: 1495956766269.jpg (157KB, 1300x892px) Image search: [Google]

157KB, 1300x892px

>>61302518
(you)

File: 1494632272584.png (19KB, 300x250px) Image search: [Google]

19KB, 300x250px

>>61304909
>but it really is more affine to pure mathematics, statistics or neuroscience
This nigga knows his shit.

File: 17800461_201105857050663_1727969891973630656_n.png (661KB, 570x960px) Image search: [Google]

661KB, 570x960px

>>61307738
>"""Pure math"""
Looks like someone doesn't know the concepts behind nonlinear optimization and how it's related to ml.

Who woulda thought the cs undergrad had no clue about machine learning???

>>61305408
>minus the (m)
Is my school the only business school that raped us with math courses? in the first two years only you take two courses in math and two in statistics. Econometrics is mandatory for almost every concentration field and some fields take no less than 5 courses in statistics and applied math.
It was good though. When I applied to a masters degree I had lots of """engineering""" classmates who knew jack shit about math, felt comfy as fuck breh.

>>61307812
You're definitely starting to shit out incoherent thoughts, m80.
And what the fuck is up with you and optimization, I'd expect a math graduate to talk about more complex stuff than that.
You don't know jack shit about math, do you? You're just a 27 year-old at the second year of a math degree in a community college, who spends his time lurking on /sci/...

What is the best way or resources to learn ML? I have background in multi-variate statistics and mathematical optimization. What else do I need? What should I focus on?

>>61307940
Andrews Ng's class on Coursera

>>61307975
What about books? Anything you recommend in particular?

>>61307938
>Optimization isn't complex
>Says the grad student who "I don't know what my focus is or how math is useful"

AHAHAHAHAHAHA

File: 1452748831427.jpg (87KB, 800x800px) Image search: [Google]

87KB, 800x800px

ya'll niggas is mad af

>>61302492
Only if you have a master's or PhD in it.

ML is easy. I don't understand why people even act like it's some abstract theoretical shit involving higher math and research. At this point, it's formulaic and ready for implementation, and most companies are just hiring self-titled "ML/AI specialists" with CS masters degrees from top rated colleges to do boring shit like model big data using RNNs to predict market sentiment or design shitty support vector machines to do complex sorting. Anyone acting like it's a big deal clearly has no clue what it's actually like in practice, or even in academia. All you really need to know is linear algebra and a little bit of matrix theory, and that's it.

>>61308194
It is the power of the marketing Jew. I mean look at Big Data™. It is just multi-variate statistics but on large datasets with the exception that you need a little more optimized algorithms to handle the large data. That is really it. I don't know what is the big deal.

>>61308219
When you consider that the IT industry is currently flooded with a bunch of scrubs who heard the dinner bell ring and scrambled to get CS degrees so they could become bottom-feeding webdevs with nice blue-collar paychecks, and most IT companies have to either hire these faggots or outsource to other much shittier countries for "skilled experts" who turn out to be even worse than the newbs, you begin to realize why this ML stuff sounds like sorcery to most people. There's just not enough people with enough experience and education to say, "oh, that's it?" and explain to everyone that ML isn't anything special.

>>61308194
Its just retarded hype surrounding it.
That and being able to figure out what values to prioritize and "teach" is harder than it sounds. But that's the job of a data scientist, not a programmer.

>>61304909
>Neuroscience

Nice meme. Neural networks have a horrible name, they don't have anything to do with a brain. What does a neuroscientist know about linear maps and activation functions?

>>61307989
http://www-bcf.usc.edu/~gareth/ISL/ISLR%20First%20Printing.pdf

http://cs.du.edu/~mitchell/mario_books/Introduction_to_Machine_Learning_-_2e_-_Ethem_Alpaydin.pdf

http://web4.cs.ucl.ac.uk/staff/D.Barber/textbook/090310.pdf

>>61306520
I'm a CS/Math double major so I think I have the authority to say that you're a fucking retard

>>61308276
>>61308309
>There's just not enough people with enough experience and education to say, "oh, that's it?" and explain to everyone that ML isn't anything special.
>Its just retarded hype surrounding it.
As a junior analyst who will enter the workforce soon I am not complaining to be honest. Let them be ignorant and pay me huge checks for doing sorcery. And yes many people tell me that what I am studying is sorcery. In my school it is the least attended field, scrubs can't into math.

>>61308343
May your bed be full of hot wives anon. This seems like very good stuff.

>graduated with Machine Learning degree.
>mfw get 300k starting

Life on easy street sure is great.

>>61308358
Go for it. I totally get that. I'm in the same way myself, got out and been working for the past few years and realized I need to go back and get a masters in it so I can put on my robe and hat and call myself a grand wizard.

File: 1487381615652.gif (984KB, 416x234px) Image search: [Google]

984KB, 416x234px

BTW if any of you faggots want to see just how easy this shit is, here's a youtube tutorial series on how to fuck around with various concepts in ML and later gets into basic neural nets using python3 and tenserflow.

https://www.youtube.com/playlist?list=PLQVvvaa0QuDfKTOs3Keq_kaG2P55YRn5v

Have fun.

>>61308343

Alpaydin's book is horrible imo. It gives very complicated explanation for very simple concepts.

Hopefully one of this days Andrew Ng will write a proper book.

>>61308062
>Literally misquoting me to fit your delusion
Stay mad little man

>>61308396
fuck, i didn't realize i will need a masters for that. my bachelor degree is Management Science/Operations Research though. But in my current masters degree is focused on natural sciences (long story, i didn't have choice on what masters to get back then). My thesis is about making a statistical model for an ongoing project which aims to reduce electricity consumption. The dataset contains around 1 million observations spread across 15 locations. I hope that will impress the employers enough.

>>61308446
I hate that nobody writes blogs or tutorials anymore. Written books are usually too dense and full of math. Practical folks now only do long ass videos.

I'd rather just read a quick tutorial with picture instead of sitting through 15+ mins long videos.

>>61308276
Is ML the only field of CS that scares scrubs away?

File: 1431559518724.jpg (24KB, 260x225px) Image search: [Google]

24KB, 260x225px

>>61308315
I cannot comprehend how intelligent your post was

I bid you a g'day, sir

>>61308532
>Written books are usually too dense and full of math.
>brainlets actually think they can compete
if you dont know the math proceed to be BTFO by everyone

>>61308194
Nah I have a friend who's doing the thesis with my same professor and they're doing pretty advanced shit with inverse RL, you can make it as complex as you like.
Anyway I agree that most of ML right now is either:
>Look! DeepMind beat X game using a literal FUCKTON of GPUs for 2 months
>Look! We can generate images lolololol

>>61308219
Unstructured data is the big deal but the buzzwords in the mainstream usage are so fuzzy nowadays that AI = ML = Data science = BI = Big Data or what I really like: cognitive systems

File: 1496263820911.gif (1MB, 800x800px) Image search: [Google]

1MB, 800x800px

>>61308758
FUCK UNSTRUCTURED DATA.
REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE

>>61308460
>Gets quoted verbatim
>Muh misquotes!!

Ok pajeet how did you score on your esl tests?

>>61304909
So should i rather study physics(has a lot of math) than cs if i want to research in said field.
Does anyone know how hard it is to get a job in IT with a physics degree?

>>61310318
>Does anyone know how hard it is to get a job in IT with a physics degree?

Literally today physics work in data analysis or machine learning.

>>61310318
Why not just study applied math

File: 1491178690839.jpg (8KB, 480x360px) Image search: [Google]

8KB, 480x360px

>>61302492
Yes. However, wait for the soon coming paradigm shift away from the spaghetti mess of convoluted statistical/math approaches that are all the rage right now.

>>61305408
What the fuck are you saying? Why are you even posting? Is this shitposting?

You just sound like a bitter math undergrad. Go fuck yourself.

>>61310239
>He disagrees with me therefore he's a nigger
Confirmed for uncultured American swine with buyer's remorse for his useless overly-specialized degree.
Go play with your autism toys and leave the grownups alone, kid.

>>61310318
This >>61310369 is true, but you'd be pointlessly working your ass off if you want to work in ML, physics is way harder.
As of now I don't think that there is a degree perfectly suitable for ML or AI, so you probably want to pick up CS and try to keep up with as much math as possible.

>>61302492
Yes. It actually requires intellect, and the bar is higher, so you won't get the occasional retard like with normal programming. No brogrammers, trendy wendies, or retarded pajeets

>>61311483
No one asked for your opinion ahmed

If anyone wants to study machine learning, go look up on google what colleges are involved in ML research, and what the entrance requirements are to get into said colleges, and then you'll know what you need to focus on in your undergrad studies.

But if you're too lazy to do that, I'll just save you the time and effort and say, major in CS, and minor in math or statistics, and you should be fine as long as you get good grades and score high on your GRE. By far the more important factor is how you plan to pay for your masters program, and who is going to write your letter of recommendation to the faculty for admission.

>>61310639
What? The probabilistic/statistical way of thinking it's the only way that actually makes some sense in the field right now;
It's way better than the shitty "empirical engineered" solutions we're seeing everywhere.
The "It just werks ™" can seem good in the short term, but in reality it's just dangerous in the long run for the field.

>>61302518
I actually like this b8 except get rid of the first line, makes it too obvious, same with the paycheck line.
Playing up the API aspect would net more (You)s.

- Tay

>>61302518
java enterprise if you want to be a codemonkey and make only 100K

If you have high IQ, learn ML/physics and you'll be making 100K+

File: 1493540777490.jpg (343KB, 640x877px) Image search: [Google]

343KB, 640x877px

>>61312857
You heard correctly. Keep your ear to the ground.
Convoluted math via overly complex statistics and compute models are utilized when you don't have a solid model for something. In leu of a solid model, the industry is using raw compute power and literally brute forcing results via distributed convergence algorithms.

A change is coming as have many fundamental changes before. See you in September.

>>61312909
>If you have high IQ, learn ML/physics and you'll be making 100K+
More accurately
>ML fags will cuck themselves out of 300k memery purely because they're too retarded for unions and are totally ok with immigrants, refugees, and visas

>Business wants to do big data
>Database is 10gb in memory

Let me offer something better. You're likely to get pulled along some CTO's vision for ML/BigData in a company that doesn't need it. You're selling something one level above snake oil, and once the ML/BigData fever wears off, you won't have a job only having managed small-as-fuck data sets.

Do this:

- Become an Java/C# enterprise dev
- Find cushy, low stress, job that lets you remote
- Maybe move to management if you aren't a complete social retard
- Kick your gf/wife out and start an investment account, unless you trust them or have a pre-nup
- Find cheap housing, maybe buy an old truck/van/RV and live out of it cheaply
- Learn to invest, esp. with options
(an decade passes)
- Do actual, interesting stuff with your life because you should be financially independent by now
- Do a few side-projects every now and then to keep your skills fresh just-in-case the market goes tits up, or build passive income products

>>61312995
And outsource

File: 1496276450114.png (133KB, 333x333px) Image search: [Google]

133KB, 333x333px

>>61313113
The wheels have been greased for something for more than snake oil. Give it some more time... Something far more elaborate lies on the near horizon that will capture minds for generations to come.