Why aren't you teaching yourself machine learning, anon?

>>60828704
It turns out, it's useless.

>>60828704
>teaching myself machine learning
Im not a machine. Delete this thread, and come back when youve rephrased your shitpost

>>60828704
It scares me.

File: 1496830959475.png (35KB, 641x527px) Image search: [Google]

35KB, 641x527px

im not a machine

YET

It's a meme.

Where's my robot waifu?

How can I get into machine learning?

>>60828724
Oh, really?

>>60828704
I've recognised each digit correctly, why would you want to overtrain me?

>>60829502
this, how do you even go about learning about machine learning

is it a programming language? how do you practice it?

it can wreck your hardware if you goof up
and your brain

>>60830163
It's ridiculously complex and too low level to be of any practical use.

>>60829502
Learn math first, at least Calc I-III, linear algebra and probability and statistics. Then look at some Coursera or w/e ML course.

>>60830193
Somewhere on /g/ an autist probably did this

>>60830706
That's far from truth, you can learn the basics and write a convolution network by hand in a week with the help of a linear algebra library. You can use it for a lot of things if you put your mind to it.

I'm too lazy to program

It's something I'd love to learn, but lack the resources to do properly. Are there any basic starting points you can share, OP?

Because I can't put it on a series of microdrones with its limited computing power it needs to be a lot bigger.

>>60831035
Thanks to you I will now take my math class seriously. I thought all this shit was stupid

>>60828704
cuz I'm dumb

>>60834039
let the machine learn for you then dumbfag

>>60834133
I'm so dumb I didn't think of that...

>>60828704
I am.

File: 1494899161616.jpg (319KB, 1080x1080px) Image search: [Google]

319KB, 1080x1080px

>>60829502
>>60830213
>>60834002
Dive right in with python 3 my friends.

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

>>60834238
Do I have to use python 3?

>>60834250
Technically no. Practically yes.

>>60834261
Does it have to do with available modules? If python 2.7 would be a pain in the ass I would rather know from the start and use 3 than use 2.7 and then find I have to rework shit into python 3.

>>60834323
Just use 3 since thats what the guy uses in the vids.

>>60834337
Here's a question which you might not know the answer to.

Why are the different python versions so separated? Why do things that work with 2.6/2.7 not work with 3? Why are people still developing 2.6/2.7 and 3 separately? Why weren't they converged?

File: xkcd_1838.png (26KB, 371x439px) Image search: [Google]

26KB, 371x439px

>>60828704
Obligatory

>>60828704
because you're supposed to teach the machine, not yourself, duh

>>60834408
Ask google.

>>60834408
Python 2:
>print x
>3/2 = 1
Python 3:
>print(x)
>3/2 = 1.5

>>60834238
thanks senpai

>>60828704
Because I know it. But I don't use it because it's a fucking hipster thing I mean I work as a C++ dev and I don't want to dive into shit with python and ML like those fuckers.

>>60834496
So in python 2 variables are statically defined while in python 3 they are dynamic state?

Who the fuck was drunk when they thought changing that was a good idea?

File: 1457147645118.jpg (7KB, 257x196px) Image search: [Google]

7KB, 257x196px

>>60834570
Totally knew you were just fishing for an argument. Bait smelled too strong so I let the other anon go after it. No regrets.

I am, I just have no idea what the hell to do with it other than feeding my autistic love of linear algebra and calculus

>>60834002
You'd be surprised how often you use, even tangentially, the skills from math courses

>>60831035
Is the Coursera cert worth anything? Am I better off just keeping a github repository with completed assignments to point at to show I know shit?

>>60834684
No. Try implementing it differently to show you can apply what you learned to different things.

>>60828704
I'm too busy with my actual job to learn something new. ):

>>60828724
>why aren't you learning statistics /g/ ?
maybe, people on this board prefer programming over statistics

>>60828704
Increase the value of every number by 1 unless it's 9, then by 2.
If the number exceeds 9, then reset to zero, likewise if it goes below 0.
ex: 9 + 2 = 1; 0 - 2 = 8.
Continue this procedure, then find the alphabetic equvalence of each numerical value, do so in pairs of 2 integers at a time. pair the numbers into double digits if it is below a value 27, if it exceeds 27, then split the 2 digits into seperate alpha-numeric equivalencies.

Transalte the images and be a amazed, for hidden within lies something astounding. It's worth it, trust me ;.)

>>60828724
It's required for some really advanced shit

It's on my list

Machine learning is such basic shit. Literally binary trees with weights attached to them. It is literally basic probability statistics. The moment you learn how it works, is the moment you realize how rather useless it is. Multilayered neural nets are much more interesting, especially training them on data sets and trying to see if there are particular neural graphs that emerge an can be reimplemented elsewhere.

>>60834496
man python 3 is awful

>>60836652
python 2: 3/2 = 1
python 3: 3/2 = 1.5
>man python 3 is awful
you fucking what

>>60835190
Neural nets are still machine learning.

But yes, other supervised machine learning methods, such as Support Vector Machines, Decision Trees, Random Forest, Bayes, etc. They are all super simple applied statistics pretty much. They can be useful for simpler problems, but they will never be perfect.

Unsupervised machine learning methods like k-nn are even worse. Again, useful for finding simple patterns, but we're not in danger of AI taking over the world.

>>60836731
How do you not see that is it pretty bad.

You are dividing integers and it returns a float? It should perform integer divison on two integers.

>>60836731
It's fundamental inconsistency that shouldn't be present across versions.

I'm trying to teach my desktop, no progress yet.

>>60830706
>ridiculously complex
As if it ever stopped people from progressing forward.
>no use
Really? Probably that's why it is used in google everywhere: video suggestions, spam detection, audio recognition, self driving cars, many search improvements. Besides google, other companies used it to identity diagnosis, predict behavior of stacks, and many other things.

Yeah, totally useless.

Sometimes I feel people on /g/ live in their own world and completely ignore reality.

How hard is it to teach machine to read my picture/scans and write a .txt for me?

>>60837937
OCR?

What's a good free trial or cheap cloud computing service that supports multiple GPU processing

>>60834250
[Code]from __future__ import * [/Code]

You're welcome

>>60828704
nice wp
what is it?

>>60838422
There are none. AWS EC2 spot instances are as cheap as you're gonna get.

>>60839377
MNIST
http://yann.lecun.com/exdb/mnist/

>>60839411
seems useful, thanks

>>60828704
But i do, I've read this book recently
http://neuralnetworksanddeeplearning.com

>>60828704
I'm not at that level yet. Haven't even done data structures