/mlds/ - Machine Learning and Data Science
>May Project
(open to suggestion, something related to /g/ traffic probably)
>4chan API
https://github.com/4chan/4chan-API
>Interesting projects, or other ML stuff, or ?
https://www.jukedeck.com
http://cs229.stanford.edu/projects2013.html - for ideas on projects, change year in URL for more
http://imgur.com/a/K4RWn - challenge images
http://vizdoom.cs.put.edu.pl/competition-cig-2016 - visual doom AI competition
https://www.kaggle.com/ - ML competitions with rewards
https://github.com/alexjc/neural-doodle
https://github.com/awentzonline/image-analogies
>Cheatsheets, infographics
http://www.datasciencecentral.com/profiles/blogs/20-data-science-r-python-excel-and-machine-learning-cheat-sheets
>Beginner Links
https://medium.com/@ageitgey/machine-learning-is-fun-80ea3ec3c471#.6wkt9refp
https://www.youtube.com/watch?v=bxe2T-V8XRs
>/mlds/ projects
>Free learning
https://www.coursera.org/learn/machine-learning/
https://www.coursera.org/learn/practical-machine-learning/
https://lagunita.stanford.edu/courses/HumanitiesSciences/StatLearning/Winter2016/about
http://www-bcf.usc.edu/~gareth/ISL/
$$$ kaggle
$$$ industry
<3<3<3 knowledge
http://pastebin.com/5a3E6Y5g - shitty R code for scraping a thread.
Monthly projects? First month, something about /g/ or /mlds/ ?
https://www.youtube.com/watch?v=qv6UVOQ0F44
>>54488458
That video is awesome
>>54488513
kinda makes you want to ML doesn't it
https://www.youtube.com/watch?v=ZeTFJfRNgqE
I think our first project should be some art person on /ic/ but I'm open to suggestion
We got some data, oh fuck I'm too high for this
but /g/ look what we got. Data mining, machine learning, sumfin, ok
пoжaлyйcтa, взять нa ceбя
https://www.youtube.com/watch?v=6wL2VvmR5MM
>>54489028
Your Russian is terrible, so is your English. Stop being high, delete that post and try writing that again.
>>54489185
https://www.youtube.com/watch?v=uEgl4Q-G6eY
Newb to ML call out, respond to this if you want to learn ML
>>54489333
yes
>>54489385
>>54488145
Caffe Berkley won't compile on my Xubuntu system. The prerec installer script won't download for some reason. How to fix.
>>54489431
Explain the bottom row. I know what I'm looking at mathematically but I still can't really visualize 4 dimensions. How am I supposed to interpret it?
>>54489463
ask /mlds/
>>54489463
It doesn't matter. Have you tried machine learning? it's easier than it sounds.
Hey, I'd love to do some collaborative work on some machine learning stuff. Whoever is the lead, please take a look at my github and see if I can be of any help.
www.github.com/chromalife
>>54489431
He's using the chain and power rules to derive the functions to make the shapes. Lrn2Calculus noob. I'm >>54489625
I'd really like in on this.
>>54489648
You should make the next OP bruv. Looking forward to working with you, a bit drunk currently.
I think this is an excellent idea. I'm really happy that others are digging it.
>>54489333
Yes, see
>>54489625
So what is machine learning an can I apply to game development somehow?
>>54489673
I'm currently enjoying a benzodiazpine so I'm not 100% concurrent. I can think of some web analytics applications while I'm a work tomorrow. I'd love to have some data science stuff on my resume.
>>54489606
Yes i have. It is much easier than it sounds.
>>54489702
https://www.youtube.com/watch?v=HePmc_XOGHI
>>54489673
Any preferred web stack? All I really know is mean.
>>54489715
I like R a lot actually.
Though I understand python tendencies.
>>54489713
Yo this makes me so happy.
>>54489702
Yes, I'm an awful drunkard. Welcome, damaged data scientists.
>>54489725
I'll learn it. Here's a list of shit I know
MEAN
Java
C++(just started)
SQL
Sadly, my shit school didn't teach python when I first enrolled, hey once you've mastered one you practically know them all.
Really looking forward to working with you weirdos
>>54489227
>>54489713
Thank you slavbro
>>54489757
> once you've mastered one you practically know them all.
This. The more exposure to languages the better.
Can we talk about Expert Systems too?
>>54489778
>tfw learned russian to read VI Arnold in native tongue
>tfw forgot it all because midwest USA friends don't care
>>54489806
sounds like a meme
We are very serious here, about MACHINE LEARNING.
What kind of Expert Systems relate to machine learning?
>>54489625, here.
I've made a copy of the links for future posts. Contact me at [email protected] so we can set up a SCRUM or AGILE meeting over some sort of VOiP. I'm headed to the third world in a few weeks to settle some estate business, but let's try to aim for a solid plan by lets say May 25th?
>>54489806
What is this? 1982?
>>54489625
>>54489875
Gross just post in the thread, friendly friend.
>RULES
>DON'T BE A GOOBER
what the fuck is this? is this skill able to be easily monetized?
>>54489919
I'm sorry sempai, benzos make me really chatty. Normally I'm like:
https://www.youtube.com/watch?v=XVekJTmtwqM
>>54489939
> is this skill able to be easily monetized?
no, nobody wants these skills. Certainly not amazon or google or anybody else.
>>54489943
my soundtrack when doing Ng's coursera
https://www.youtube.com/watch?v=OrMjc4oz0mY
>>54489951
okay great, i will ignore this then :^)
>>54489968
Was never really a fan of my bloody valentine. But this album is how I got through systems analysis and design this semester.
https://www.youtube.com/watch?v=mmxBoFzdqVI
>>54489991
cool
Hey what should our first project be? analyzing /g/ in some way is my vote.
>>54490024
I mean making something simple like a webcrawler to see if companies really do shill on this site would be pretty neat imo. Also the sequence of boards that people visit could be cool as well. I've loads of ideas.
>>54490049
> I've loads of ideas.
spill'm. That's what we're here for bruvfam.
We could make a doomed-to-fail kaggle team, a github team, whatever else. It's about time I make a new, anon github. Hoping for delicious PRs.
I think we should agree on diverting money in the unlikely case we win anything. EFF?
>>54490049
How would you achieve any of that?
>>54489606
Yes. You should really add NN's demystified to your op.
>>54490084
Well you could use a TreeSet or HashMap and store IP's and the board urls as nodes on the maps. Each IP would have a uniqueness and then you feed that raw data to the algorithm. This is pseudo-code of course so this may not be the "correct" implementation, but it could work in theory.
>>54490105
>our op
suggest shit, make a pastebin, I was just throwing shit out my bookmarks. I'm a drunk failure, I just want this to happen so I can learn from smarter people.
https://www.youtube.com/watch?v=L53gjP-TtGE
>>54490122
My point is how are you planning to get the IP's?
>>54490123
Here's the old op I used to use for it. Good luck and godspeed. I'll be watching.
>No /aig/
Artificial Intelligence General - FUCK ALL YALL edition
Come and discuss, learn about, and panic over all things artificial and (marginally) intelligent.
Links:
>Deep learning frameworks
https://www.tensorflow.org/ (google)
http://caffe.berkeleyvision.org/ (Berkeley Vision and Learning Center)
http://deeplearning.net/software/theano/ (Université de Montréal)
http://www.cntk.ai/ (Microsoft)
http://torch.ch/
>Frameworks comparison
https://en.wikipedia.org/wiki/Comparison_of_deep_learning_software
>Required reading
http://www.deeplearningbook.org/
>Intro videos for beginners:
https://www.youtube.com/channel/UC9OeZkIwhzfv-_Cb7fCikLQ/videos
>Required watching
https://www.youtube.com/watch?v=bxe2T-V8XRs&list=PL77aoaxdgEVDrHoFOMKTjDdsa0p9iVtsR
>Something else that you should probably watch
https://www.youtube.com/watch?v=S_6SU2djoAU&list=PLy4Cxv8UM-bXrPT9-ay4E1MuDj1KFTg9H
https://www.youtube.com/watch?v=6oe1Tmg9rjM
>Books
magnet:?xt=urn:btih:a680ca553dd69a6a5e8b7dc8c684ceb006c7ecc5&dn=Artificial%20Intelligence
>Other Resources
https://facedetection.com/
http://opencv.org/
http://kaldi-asr.org/
http://sirius.clarity-lab.org/
How are those networks comin' along?
>>54490136
That's a good point. I dunno actually. We could use cookies instead.
>>54490151
Same problem, there is no way we will get access to any of that.
>>54490148
I remember this. I just got done with the semester so I want to do a lot of learning/projects over the summer. We should start an irc or all just join #/g/sicp on rizon.
Very interesting stuff.
I'm having problems getting into that last.attempt email address. I'll post a new thread when I get to work tomorrow with another email for those who are interested.
>>54490179
Hmmmmmm....
Perhaps we could ask the clover dev if we could use his code for analytics purposes?
>>54490183
Here 1 sec.
>>54490148
http://pastebin.com/Ta697eTi
Now let's settle some shit.
My opinion: industry/academics is into ML and "data science" (essentially applied statistics)
I think we should go with /mlds/ - ML & DS
So now categories for an OP
I don't know, I think we should focus on ML. There's enough out there to keep it interesting and we can come up with new great stuff, because it's new, and industry's thirsty. Plus ML applies to AI, and AI would be welcome. But imo the general should be ML and a side of data science/AI.
>>54490200
Yeah, because why wouldn't anyone agree to that?
It's not like this website is considered to be one of the few bastions of internet freedom left. And it's not like there were shitstorms when Hiroshima came to power because people thought he would use/sell the data collected here.
>>54490151
>>54490179
You can do a lot of interesting analytics with time, comment content, attachment, (trip maybe, especially if we have a project we trip for)
IPs/geolocation is a whole other thing, it's better that 4chan is anonymous in that respect
>>54490229
>sarcasm detected
I forgot that it was open source okay? It's been a hellish few days for me.
>>54490183
https://discord.gg/010Hh6pWkFkahZ8yM
Made a discord server for it.
>>54490249
>requires javascript
okay or we could talk here
>>54490244
Agreed. Maybe do something like see the peak hours for which kind of threads?
>>54490249
>>54490261
Lol yeah, I'm on my ubuntu machine, dunno if supported.
>>54490248
No bully intended, just pointing out that we have very little to work with. Also
>I forgot that it was open source okay?
This triggers me, we are not microshaft, let's not go the evil root.
>>54490262
Yes! All kinds of possibilities. First things is getting clean data. I do R but that's just me. I pastebin'd my shitty code for scraping somewhere but I'm heading off now.
>>54490285
I've no R experience, but I'll just watch some lynda videos on it. I'm excited!
>>54490277
The web app is supported on ubuntu
>>54490261
really?
>>54489701
AI.
>>54490296
coursera's "data science specialization' uses R. It's all free, also a lot of the assignments are using this library "swirl" that anyone can install and use.
So basically >install.package(swirl); >swirl() should get you into it, it's kind of a hands-on way of learning. I don't know I'm explaining it badly just get into coursera and do "R Programming" and it's clear.
I wrote this little thing that applies gradient descent to find the regressors of a series of points. It's probably the simplest form of neural network possible.
Where do I go now? I kind of want to learn how to make each kind of system so I really understand how things work by myself but even not too complicated stuff seems like it would take me a week to complete.
What would be a nice roadmap for increasingly harder (but more interesting as the difficulty goes up) projects?
My goal would be to make something like that car that drives itself just by analyzing a black&white image of the front view to steer itself properly. I'd want to implement that in a game and see the AI grow.
>>54490329
I'm partial to Lynda personally, but if I run into roadbocks ill use coursera
>>54490347
That's over my degree specifications sadly. I have code monkey degree.
>>54490347
I am looking forward to checking this out when I'm not drunk just in case nobody is interested. I probably won't have mucch to say then though either but I'm interested.
I have a background in finance with ok starter statistics (arima/box jenkins, VAR, ...). I'm no programmer but I can code and learn enough to usually do what I want. Would getting a job in ML be possible? I think I have enough of the stats wizard skills but I fear my programming skills are not impressive enough to work on a team.
>>54490377
It's not really the stats that would be an issue for you. The linear algebra and derivations would be the hard part.
We could write an extension? For Chrome/FF?
>>54490391
I've followed Andrew Ng's course up to 80% and could follow pretty decently. Does it get much worse?
Not relates but I was very interested in the NEAT algorithm which mixes neural nets with neural evolution to find the best solution. That system just blew me away when I first read about it. It's fantastic, but I couldn't get it to work in a project of mine.
>>54490391
https://www.youtube.com/watch?v=ZK3O402wf1c&list=PLE7DDD91010BC51F8
>>54490415
Fellow member of the Dr Strang Love club.
Reminder that his book on diff eq is published and videos coming out soon.
>>54490414
>Does it get much worse?
Nah, I mean, not so far. It's kind of a new field. But equipped with calc and linalg and Ng's coursera, you are in a position to do some seriously awesome stuff.
>>54490489
I see. The thing is I've followed 70/80% of Ng's course and while I get how to sort do logistic and regular regressions, I don't feel like I can really apply this to any problem of interest to me.That's kind of the main reason why I stopped the course.
I guess the right thing to do would be to finish that course I suppose.
>>54490502
> while I get how to sort do logistic and regular regressions, I don't feel like I can really apply this to any problem of interest to m
Sounds like this general is perfect for you, and for me, since I feel the same.
>finish that course I suppose.
Also maybe set up an anon github, and I'll do the same, and probably others will do the same if they haven't already, and we can pull request eachother lustfully.
>>54490529void kek (int * chikn, float wattermellun)
{
nigga.muffugin(bix nood);
for (money_fo_dem_programs = nood; money_fo_dem_programs < infinity; money_fo_dem_progrmas++)
{
cout << "Dindu nuffin' man" << endl;
}
if (you == up)
{
smash(muffugin_like_button);
}
}
>>54490620
Cool! Great effort here!
>>54490415
This is an essential playlist.
Here's another:
https://www.youtube.com/watch?v=j9WZyLZCBzs
Lighting's a bit fucked but it's not a dealbreaker if you're into it, he's a really good lecturer.
So I wanted to test a bunch of subjects in a high variable environment, such as drivers on a freeway with accurate collision avoidance etc. Am I looking for ML or maybe a genetic algorithm? Looking to find the best fit driver within the system considering multiple variables.
TensorFlow, BigQuery, BIG MONEY, BIG RAD
https://www.youtube.com/watch?v=cp8gjhWaPjw
