Machine learning

Why aren't you learning machine learning?

Other urls found in this thread:

youtube.com/watch?v=TOsMcgIK95k&index=1&list=PLbtzT1TYeoMjNOGEiaRmm_vMIwUAidnQz
dkriesel.com/_media/science/neuronalenetze-en-zeta2-2col-dkrieselcom.pdf
github.com/HFTrader/DeepLearningBook/blob/master/DeepLearningBook.pdf
deeplearning.net/tutorial/deeplearning.pdf
neuralnetworksanddeeplearning.com/index.html
twitter.com/AnonBabble

Becaue the machine should be learning. Not me.

>2010+6
>not being a machine

lmao biofags

They are efficient desu

>1999 + 17
>not being an augmented human

literally doing this rite nao!

>666 + 666 + 666 + 6 + 6 + 6
>not being a machine-made human whose purpose is to learn like a machine

Should I?
It comes with a lot of math courses.

What frameworks would you suggest?

Is tensorflow the best way to start? Or is it better to get the theory down first

what i like about machines and computers are their predictability, machine learning literally kills what i like about technolgy... how should i feel?

That's what I'm trying to wrap my head around right now.

First I have to understand computability. I found lots of great resources online.

This guy has a good series on computability which is very understandable and thorough:

youtube.com/watch?v=TOsMcgIK95k&index=1&list=PLbtzT1TYeoMjNOGEiaRmm_vMIwUAidnQz

And here is a good non-technical introduction to neural nets:

dkriesel.com/_media/science/neuronalenetze-en-zeta2-2col-dkrieselcom.pdf

And a more thorough text on deep learning:

github.com/HFTrader/DeepLearningBook/blob/master/DeepLearningBook.pdf

And finally a more thorough text on various actual code implementations of deep learning:

deeplearning.net/tutorial/deeplearning.pdf

We aren't so much programming things as breeding them now. We "train" them as well.

>666 + 666 + 666 + 6 + 6 + 6
what the actual fuck? I am scared now.

I'm currently doing machine learning in NLP. Trying to get out.

Fucking machine learning is all about quality management and win:loss ratios. You tweak the parameters, choose a different weighing function, and see if you get a 2% relative improvement over the status quo.

The scikit-learn (sklearn) module for python is pretty damn saucy. Like most python, it's dial an algorithm and is very easy to implement some serious shit.

If you want >9000 computing power and have a spare NVIDIA card lying around CUDA is also awesome. Though you'll obviously have to write your own/download some one else's AI code. Great for evolutionary algorithms and neural networks.

I'm seeing a lot of nu-males doing machine learning. That's reason enough for me to stay out.

Go dig ditches then, alpha man.

thanks a lot for the links

but where do i even start?

Go to a proper university to study computer science for four years.

Then apply to a good grad school that has a good research program in place for machine learning.

autistic

Most publishing occurs at business firms now for this sort of stuff, AI is hot again

Implement by hand first tbhfam

Good framework depends on what language you want to use.

Python is the most common language used. I haven't personally used it but I heard Blocks is good.

>tfw it will die again after researchers fail to pump out revenue-generating product within 24 months
>tfw 24 months is the maximum patience of most businesses

Because it's a meme.

And I am not gonna fall for a meme.

Really makes you drink...

and what if i dont want to dedicate my life to formal research? and just do it as a side project or something to tinker with

wtf I hate 2016 now

because [spoiler]I'm dumb at programming ;_; [/spoiler]

But I'm already learning, user
Started reading this last month and I'm almost finishing it.
Really simple explanation, without much of the complicated math (although it's probably a good idea reading a bit on the math side after familiarizing with everything)
neuralnetworksanddeeplearning.com/index.html

Thanks for the links

Because I haven't been able to find any good resources online. I've only Googled a badly worded term once though, so no real effort put in. If anyone has good material I'd love to hear (already noted down the links of earlier anons).

Wow, yet another thread confounding the fucking massive field of machine learning with:

- Deep learning
- High performance computing
- Prodigious programming and math

Tip: If you're at all interested in applied ML, literally just read / take basically any non-entry level stats or business analytics book / class. There's a lot to learn (and none of it will happen on Sup Forums), but contrary to the popular advice given in these threads, you don't need to be a master of any field to get started learning. (But on the same token, realize that "muh leet programming skills" aren't going to get you very far either).

I'm doing linear algebra next semester, so maybe I'll start it after that.

I can't do calculus.

it's 2016, no one can, just use wolfram alpha

kek

>tfw my dream is to make a contribution in the field of AI or ML
>tfw i was born a fucking idiot with low self esteem.
>tfw shit programmer

i wanna die

I hear Microsoft is looking for people just like you.

no one is looking for people like me.

Are you gonna cry

That is good.
You know who is being looked for? Escaped convicts.
Do you know who isn't being looked for? Ninjas.

Would you rather be a convict, or a ninja?

i already did that. i'm probably just gonna go to sleep

are you that fat girl from middle school

You should probably understand computability and some set theory

holy shit mane

bumpu

saved this qt image woah

Because I have already learnt it and now I'm teaching my agent to play a videogame.
Stay pleb, Sup Forums

>tfw false highs and true lows