/sci/ - Science & Math

>>9740328
>B.A. in Business Strategy
>John Hopkins University Certified Data Scientist
This new age bullshit is getting out of hand

>>9740328
PostGrad in Data Science: no, it's not.
It's just a new name for statistics and applied maths. The "collateral" training you get in CS is easily achievable by forcing yourself writing some code.

> There is no such thing as Machine Learning
> There is no such thing as "ai"
> There is no such thing as "Data Science"

For most the algorithms everything boils up to a model where you try to minimize the error. That's it.

>>9740357
This. It's looking at time series graphs + regression.

>>9740359
Pretty much. Most complicated thing I have seen was (linearly) combining PDE with spline regression.

Also, do not fall for the Neural Networks meme. It's literally: "we need a big dataset and a large number of slave wagers to randomly fit tons of model until the best result is archived".

>>9740328
statistics is theoretical data science

convince me otherwise

Well, given that chemistry is considered science, I don't see why data science or gender studies wouldn't.

Yes, data science is a thing
sorry for the non-autistic answer. I don't even know whats going on in this thread right now

Of course it's a thing, it is just often called differently.
In it's core, it's basically statistics, although depending on your work location and your field, it can vary what other qualifications you need as well.

We have a data scientist in our institute (Chem), who helps with data analysis for research etc.

>>9740328
Yeah, it's just statistics applied to data.

>>9740357
>PostGrad in Data Science
How do you like it so far? How is the market for data scientists right now?

>>9740328
>>9740339
>>9740357
>>9740359
>>9740366
(you)

>>9740328
"Certified data scientist". What does that mean because it sounds like some bullshit.

>>9740394
Pretty saturated, but also lots of demand, so if you are not a retarded you have good chances of finding a (boring) job where they probably expect you to care about everything data related.
Personally got 3 interviews, but 1 job offer in bank and a phd studentship, which I'll take.
Also, I'm not an English native speaker, and I am terrible at interviews. I can tell that friend of mine managed to get 3 offers and a phd, though. If you do not have ambitions in life, working on data that are definitely messy and where you'll have to explain retarded-level distance functions to excel-based managers, then go for a job in Data Science.
But if you want to have fun in researching new stuff, making new models and develop new techniques, go for statistics or mathematics: you'll get a wider range of tools for playing around (prepare to live on grands tho).

>>9740400
It is, probably someone pulled out a course where you pay your 3k and you get a certification for "ahah I know how to import numpy on jupiter" level of skills.

>>9740406
>pip install numpy
>Here's your certification anon

>>9740408
Typed that in word, does nothing, I'm stuck help

>>9740406
>If you do not have ambitions in life
I assume you are talking strictly academic ambitions, and not monetary?

>>9740406
>I know how to import numpy on jupiter
The hard part of imporing numpy on jupiter is getting to jupiter on NASA's budget.

>>9740328
The Johns Hopkins DS certification is an online certification that you can get through Coursera.

>Forbes
>SAP
She's obviously just a business college graduate who did some certifications so she can be somewhat qualified for working with a business team that deals with data and analytics.

Nothing wrong with that, but it's not that related to Math or Science..

>>9740408
Lol. (If using windows) You need to go to your python install and go to the scripts directory and see if pip is installed. If not you install pip for your version of python.

Next open cmd and cd to the python scripts directory, from that directory run pip3 install <package> if on python3 and pip install for python2

If you're on ubuntu or something you can use apt-get install

>>9740898
Nothing wrong with her qualifications but it's how she uses them.

>>9740328
It has a very big future, anon, unlike most math graduates :)

>>9740328
I’m a business undergrad. My school offers these “data analytics” certificates as like an extra three classes you take on top of your major. It’s basically just learning how to use various database and statistical software. No math classes, moreso just skill training to pad your resume.

>>9740966
Do they teach Python as well?

>>9740923
Look at all those equations on the top of her Twitter! She must be like a master mathematician.
Can’t really blame her though, that is exactly what they tell you to do in business school. I had to take a whole class about “marketing” myself I.e. distilling my personality and experiences into whatever buzzwords employers are currently looking for. It makes my stomach turn.

>>9740970
Maybe I’m not that familiar with it. I have zero math/computer knowledge whatsoever and don’t know why I post here.

>>9740914
Either that or just download Anaconda and get python + a shit ton of packages.

Data Scientists are pretty much just doing linear regression and some K-Means clustering.

Anything beyond that is too "researchy" for practical use in the business world. If you can't easily explain it to your manager, they aren't gonna let you waste time working on it.

>>9740956
I'd hire a pure mathematician with statistics knowledge over a data scientist or statistician 100% of the time.

>>9741137
id hire someone who won a kaggle competition

>>9741137
As long as the pure mathematician knows how to code and deal with Big Data using Hadoop or whatever and use Machine Learning/Deep Learning models, then I will agree.

>>9741171
>Hadoop
the biggest fucking meme in the industry

the number of fresh grads who have hadoop on their resume as if it is some badge of intelligence is mind boggling.

>>9740357
>The "collateral" training you get in CS is easily achievable by forcing yourself writing some code.

No, traditional CS studies is typically: theory of computation, database theory/applications, computer organizations, operating systems, algorithms, discrete mathematics, probability & statistics, linear algebra, calculus I-III, and then 5 or so upper division CS courses like cryptography (which at my school is all pure math based), machine learning (heavily linear algebra + stats/probability based), computer vision, NLP, advanced versions of ML, computer vision, and NLP courses (all separate courses) and advanced versions of algorithms, theory of computation, computer algebra, etc courses.

Learning code monkey skills doesn't make you a computer scientist.

>>9742099
also, at my school CS majors can take the theoretical linear algebra course over the applied one. There are CS students taking courses in real analysis, abstract algebra, topology etc as well

>>9740357
>There is no such thing as Machine Learning.

What about reinforcement learning?

>Machine learning algorithms just try to minimize error.
What if, along with reinforcement learning, thats all the human brain does?

>>9740357
This. Basically statistical analysis of some database. Data Science is just people who maintain all the data mining crap everyone is crying about that Facebook/amazon/google are doing. Well everyone wants in on it and the big guys keep it all proprietary so start ups need people who can write just enough code to plug in statistical correlation tests to their data

>>9740328
>DATA SCIENTIST
>GROWTH HACKER
>TECHNOLOGIST
>Good decisions are made from good data. Good data is made from people you can trust. Sarah trusts herself.
cringe/10 what a fucking joke

>>9742201
lmao

>>9740328
It says right on her pic there that she's a marketing floozie.

>>9740339
I'm a certified wallet inspector

>>9740328
>John Hopkins University Certified Data Scientist
I did that course, all it is is basic R programming (I think they switched to python later idfk) and very basic statistics. Also includes some aux skills that you can easily learn yourself like how to document your work, and how to share it with people. Its 9 modules and I spent about 1h on each except for the last one which is a bit longer.

data science isn't real.
a data scientist is a statistician who isn't shit at programming and is capable of operating professionally without a middle manager constantly spoonfeeding them bullshit tasks and holding their hand

>>9740328
A woman's mathematical ability is inverse to her attractiveness.

In her case, she's probably ok at maths.

>>9740381
welcome to /sci/

>>9740555
badum-chh

>>9742483
>a data scientist is someone who uses R and python

Is there anything less sincere than a person advertising themselves with formulas in the background?

Isn't there a more tasteful way of doing this?

>>9742556
it's even worse when some of the formulas either make no sense or are blatantly false

>>9740328
It's basically statistics with some basic programming skills. The reason people call it data science instead of statistics is because it's a huge buzzword with employers who don't know dick about dick.

>>9741153
The fuck is that?

>>9740328
How can you be 30 if you're under 30?

>>9742570
Not that guy but Kaggle is a site that hosts a lot of machine learning competitions. They attract a lot of people and the winner is whoever can make a classifier that outperforms all the others. From my limited experience the people who win these things not only have to know what they're doing on the machine learning side of things but they have to have some amount of domain knowledge and just be pretty bright overall.

>>9742575
Ayy lmao

>>9740328
>http://sarahaustin.com/growth-hacking/data-science/
I vomited in my mouth when reading that website. She has some technical background, probably even visited some statistics classes, but that's all the positive points.

I see people like her all the time. People who think they're the next cool thing, surrounding themselves with a cloud of the latest buzzwords and hip terms. Of course, she is an influencer, storyteller and a a growth hacker. I wouldn't expect anything else.

You want to know what the reality of working with people like her is, however?
Most of the time they are a complete fraud.

She preys on people and companies that don't have their business completely in their own grip. She make you believe that she's some sort of savior with mystical powers, "you give me your data and i will magically save your business". For her unknown mystical services (she will evade your attempts at making her give you an exact definition) she will demand ridiculous fees while skillfully avoiding any responsibility or KPI you'd want to impose on her. She will be prepared to blame anyone except herself for the eventual lack of success - other partner companies or agencies that work for you are usually the easiest target.

The first three sentences are all you need to know about her. A real "data scientist" would present their results and case case studies about past clients. "I worked with X, analyzed Y, discovered A, B and C. We've prepared strategy D, the results are E% increase in sales and etc." Her presentation is vague at best.

>>9742572
Lol. It's a Forbes list of 30 people are under 30.

>>9742483
This is the only correct post ITT

>>9741137
But then you’d have to pay them $300k a year.

>>9740388
Statistics is already just probability and combinatorics applied to data.

>>9742556
Tasteful? That’s Twitter.

>>9740914
>Next open cmd
You need to open it as an admin
>If you're on ubuntu or something you can use apt-get install
You forgot sudo

>>9743074
> He does not run everything as root

>>9740381
/sci/ = worst!

>>9741801
What would you recommend then?
Is SQL, .net, and python a good bread and butter then?

>>9741801
I take it you've never worked with a dataset bigger than a few dozen GB.

i picked a random article she wrote and went to a random sentence and this is what i got
>In my own work, I enjoy creating computer programs that are based not in math but in pattern recognition.
https://www.entrepreneur.com/article/295460

>>9743300
>being this much of a brainlet

it's a thing because people let it be a thing, and people hire for that thing.

Basically, the role is more math/stats than your typical programmer, and more programming than your typical math/stats person. And some sense of presentation/design/visualization.

The reason it didn't disappear immediately as a buzzword, is because as much as it seems like you can chuck everything into the ML black box, it's a little more nuanced than that, and average programmers don't have the stats and math background, and average math/stats people are autistic.

>>9741153
But isn't that like hiring a basketball player based on their achievements on a slam dunk/free throw competition as opposed to their performance during official games?

>>9742575
Ayy lmao

>>9743654
>SQL
Yes you should be familiar with it.
>Python
Also R. Unless you want to piss off IT, R is a must. For example, does scikit-learn support MARS?
>.net
Learn the Linux command line. Don't waste your time with this.

>>9746739
this. R is always going to be a step in front in this area, because statistical computing is its exact problem domain. Python has a lot of "data science" tools and libraries, but it's also a lot more general purpose. Learn it for sure, but don't imagine it replaces R.
Simply the fact that CRAN is so wonderfully curated, goes a very long way to keeping R important. You'll find researchers putting out an R package far before it comes to python.

>>9740328
data science certification is a complete meme
a graduate degree in data science involves taking programming classes, machine learning classes, statistics classes, optimization classes, combinatorics and graph theory classes, and some applied math classes.

>>9740328
call it automated statistics

>>9742590
Let up dude. She's a young college grad who went to BUSINESS school. This is literally what they teach you to do and this is literally what works in our dipshit society that only cares about checking boxes. She's literally just acting in her own interests, just like anyone else is

Itt, /sci/ discovers data science

I can't wait for all the data science memes

>>9740328
30 Under 30 is the hallmark of bullshit. A few people I used to work with got it, journalists just put their SJW friends on it because YASSS WOMEN IN TECH

CV goes straight in the bin as far as I'm concerned

And yes, data science is a big meme

>>9746739
>R

Literally why

Numpy+Pandas is more than sufficient

>>9740328
I'm getting way too angry looking at this picture.
Fuck these retarded meme fields, this is why women shouldn't be allowed to work or use computers.

>>9742590
>For her unknown mystical services (she will evade your attempts at making her give you an exact definition) she will demand ridiculous fees while skillfully avoiding any responsibility or KPI you'd want to impose on her. She will be prepared to blame anyone except herself for the eventual lack of success - other partner companies or agencies that work for you are usually the easiest target.

Like every bussiness student.

>>9746739
>>9746759
lmfao, idiots. R is a shit programming language only used by plebs with datasets that fit in memory. anyone with a non-trivial amount of data is probably going to be using python, java, or scala

btw, there are tons of open source implementations of MARS, even for sklearn
https://github.com/scikit-learn-contrib/py-earth

>>9748890
the graphical abilities negro

>>9748890
Shitter spotted

>>9748890
>using python for statistics
absolutely disgusting

>>9740339
>ba
>arts
Business strat is an art?

>>9742575
ayy lmao

>>9740328
>B.A. in Business strategy
What the fuck does this even mean? How do you get a degree in this and what do you even learn?

>>9740357
This. Another knowledge bomb:

>There's no such thing as multiplication
It just boils down to repeated addition

>>9741137
And which company people should care about do you own?

>>9743667
Not OP, but Hadoop is a shill unless you're working with greater-than-Terabyte level datasets. Most data scientists I know use Spark because holy shit Hadoop is slow. If you're Google-scale and have terabyte or petabyte-level jobs then maybe Hadoop/BigTable makes more sense.

>>9748890
lmaooo good luck plotting literally fucking anything

>>9748962
nobody tell this dude that Python's premier data analysis library (pandas) also works in-memory.

In reality, if you're working with non-trivial amounts of data you're using either a database (SQL) or some binding to Spark/Hadoop in your language of choice (including R/Python/Scala). But nice try, retard.

>>9750152
How bout this knowledge bomb:
>There's no such thing as addition
It just boils down to counting

>>9750442
Actually there’s no such thing as multiplication. It’s just shorthand addition.

>>9750297
>he doesn't know what out-of-core learning is
>he doesn't know about blaze

try doing some out-of-core learning with R, you little faggot. literally every SO and Quora post asking how has the top reply "just buy more RAM dood"

>>9742575
AYY LMAO

>>9748962
>a non-trivial amount of data
so this is the power of data science

>>9740328
>B.A business strategy
probably has a minor in synergy
Why are these people allowed to exist.

It's just the "science" of information which is actually what statistics is.
Just looking at the information in the way that gives you a useful answer.

wait to people here really don't know what a data analyst/scientist does?
I must be missing some popular meme here there's no way you guys are so out of it

>>9752057
data scientists are the meme, my friend

>>9753086
but they're pretty much just statisticians what's wrong with you my man

>>9753124
the OP girl is not a statistician. she's a clueless business major with a cringe tier website who probably couldn't write fizzbuzz

>>9753127
?!
I wasn't talking about whoever it is in the OP I was talking about data science as a profession
but never mind, I just realized why I don't actually come here so we can leave it at that
btw the code in her presentation isn't even hers, if you google for it you'll find it on plenty of other websites used as an example

>>9753128
and yet you're here. the meme part comes in when people as clueless as her know enough to pass as data scientist

>>9750448
>addition.
You mean iterated successor function?

>>9753130
>job X is a meme because there's people out there who are shit at interviewing candidates
oh no!

>be accountant who can into vba, sql
>company falls for the big data meme
>hires me as "senior business intelligence analyst"
>no one asks for anything
>make $80k a year f5ing 4chan all day

>>9753263
I'm the opposite

>learn python and sql in my first job
>second job get basically the same as you (BI data analyst meme job)
>about $75k a year
>either have nothing to do or completely swamped, no in-between
>absolutely hate it

I am literally a wageslave

>>9753263
anon that's great!
unless you're a woman, then it's awful and evil!

>>9754281
>either have nothing to do or completely swamped
welcome to project-based work

>>9754332
Oh nononononono

Data Memeing is never project-based

It is always ad-hoc tasks from management or salespeople, never anything big or meaningful

I do know Data Scientists who HAVE been assigned big projects, usually what happens is
>request is vague and not really possible ("we want to do our ads with machine learning")
>request is straightforward analysis which is then ignored in favour of the manager's gut feeling on the issue

You will not contribute to anything in the company in any meaningful way, you will always be either reporting monkey or (if you have a PhD) a name they can use to impress investors with

>>9740328
you should be ashamed of yourself for making this thread.

>>9754411
or y'know, actually find a nice company to work for and have meaningful contributions and collaboration
while GDPR will likely have nasty implications for data science, I think just about any big-enough company that has IT/web services as its core business can be a nice play to work as a data scientist/analyst, or even any company that makes data driven decisions
if what you said is the kind of work you do then that's obviously not a good place to work at

however!
>request is vague and not really possible
if the requester isn't a complete idiot they will be open to communication, on average they themselves don't really know what they want let alone need, and that's where you, as a capable engineer, come in, take them to the side and talk to them and together try to understand what it is that you're all trying to do and how to go about it
>analysis is ignored and managers go yolo
see previous point about good places to work; if that's really the case then you can half-ass your work and use that time to study other things that will land you a better job somewhere else

but again I think in these cases you only have the title of data scientist, not the profession; just like an account manager isn't an "actual" manager, just like tons of engineers aren't actual engineers; just like a devops engineer can't even fucking exist because devops isn't a job title, just like everyone saying they do agile doesn't mean they actually do agile
you can slap whatever label you want on whatever the fuck you want, don't undermine the people who actually do it just because it's misrepresented by fucktards

Data science is frustrating when everything you do is controlled by business people with 0 statistical experience.
I was asked to predict some outcome, but they couldn't get me any data on whether or not the outcome actually occured. So I just present shitty hand-wavy algorithms and say "I think if these things happen, the outcome will probably happen too".

>>9740357
>For most the algorithms everything boils up to a model where you try to minimize the error. That's it.
this is how I know /sci/ is shit at stats

>>9754432
>or y'know, actually find a nice company to work for and have meaningful contributions and collaboration

I'm not saying these companies are bad to work for, I'm saying that 90% of the time when they say they want a "Data Scientist®" what they actually want is "Reports bitch" or "PhD for marketing material". Same thing with Quants in banks.

>you can slap whatever label you want on whatever the fuck you want, don't undermine the people who actually do it just because it's misrepresented by fucktards

Fair enough, the pure DS guys I've worked with have all been solid, intelligent guys with a defined skillset, just never liked seeing them get hired and then having their expectations of the work completely dashed

>>9754439
>I was asked to predict some outcome, but they couldn't get me any data on whether or not the outcome actually occured. So I just present shitty hand-wavy algorithms and say "I think if these things happen, the outcome will probably happen too".

Sometimes that's actually what they want though

The other times they'll take your approximate estimate as gospel and attack you when it doesn't play out word for word

>"I want to be a data scientist!"
oh cool. what kind of data sciency stuff are you into?
>"Big Data!"
Oh nice... so like clustering? Unsupervised learning?
>"AI!"
Oh? So like reasoning and Planning?
>"DEEP LEARNING"
Oh I see. So you are into Computer vision, or maybe NLP?
>"N E U R A L N E T W O R K S"
right....
>"Actually I just do linear regressions in excel"
Oh....

>>9754753
I am a data scientist. Big Data. Big Data = Big Brain. TensorFlow and Keras are my specialty, but I'm very well trained in 'for loops' and searching stackexchange.

I cannot write the definition of a derivative nor can I provide an intuitive explanation of the derivative of a univariate function.

>>9740328
Is this the golden road to data science?

>>9755769
the average 'data scientist' has not heard the words Monte Carlo in their life

>>9740328

People can bash it

But I'm enjoying my over inflated 100k postgrad salary

Enjoy your 40k postdocs

>>9748888
in what way is data science a meme
it seems like people on /sci/ have misunderstood what it is

>>9755769
>"Working in Excel" is the 2nd to last in "programming"

>>9755774
They sure as hell have, where else would someone with 100k starting salary spend their vacations?

>>9740328
Sort of. Data science could be considered as applied mathmatics on data sets, particularly large ones. This tends to overlap with mathmatical modeling and database/programing knowledge. You can come up with hypothesis about the data and then test different databases to see if it's accurate.

For example, someone could say San Francisco produces more air pollution LA. If you have data samples from both locations you could analyze the data to,get average amounts or other metrics to compare them and judge trends. In these cases I don't consider this far from a scientist testing hypothesis about food intake or drug treatments. They also are just doing sampling and stats to support a hypothesis.

>>9755774
Maybe they've been there though

>>9758830
Pretty sure that it's the 2nd thing in programming in the picture made by some retard.

>>9756935
>Enjoy your 40k postdocs
will do brainlet