Explain Machine Learning to me as if I'm 13-years-old

Explain Machine Learning to me as if I'm 13-years-old.

Attached: GOOFY.jpg (240x240, 7K)

Other urls found in this thread:

youtube.com/watch?v=aircAruvnKk
twitter.com/SFWRedditVideos

Underage b&

Okay, OP. You have a deal.

>as if

rotate vectors until you see shit

cya in 1825 days

Imagine you are in a shop and you want to buy your nephew a toy. He is a 6 yo boy. So you think: "1. He is a boy. 2. He is a kid. His mother told me he likes 3. cartoons". What would you buy? Would a pink purse be a nice gift? Probably not. But why? Because as we know he is a male kid and from our experience we know that these are not things that male kids like. Computer has no experience but we can give him information about kids and gifts that they liked or disliked to teach him about it. Computer can decide pretty easily. It looks at the all the information that you have given to him and makes decisions between them. It looks something like this: (boy or girl) -> boy, (older than 7 or younger than 7) -> younger, (older than 3 or younger than 3) -> older, (likes cartoons or likes dolls) -> cartoons. From his [computer] experience kids with such attributes liked big dinosaurs the most.

Wow that's pretty dumb. Why not just ask the boy what he wants and tell it to the computer?

>he is a boy
>male kid
>boy or girl
TIL machine learning is sexist and is enforcing harmful gender stereotypes

Attached: 7dqZSav1.jpg (480x480, 24K)

Attached: machine_learning.png (371x439, 28K)

The point ml is to predict things. What's the point to use it if you already know the answer.

Why does a computer need to predict things? This is just a waste of time, time which could've been spent on finding a cure for cancer. These scientists don't have their priorities straight.

because people hate being interviewed and if i want to farm for answers ill interpret your activity on the internet instead of explicitly forcing you to give me the answers

once the answers are there the system just uses statistics to extrapolate your next most probable action

Cancer curing is one of the reasons it is so popular. It can predict what treatment is best for you if it knows what type of tumor you have and its size

Over-glorified pattern recognition.

>pattern recognition
thats exactly what ML is you stupid fucking faggot

stick to desktop threads and stop using words like over-glorified in hopes that someone will take you seriously

Imagine a box that takes an input, and an output. Inside this box, is maths that simulates the basic understanding we have of brains, neural connections. We force data through this box, and the maths does it's work and provides an output. But, at the start, the output probably won't be anything close to what we expect. We keep forcing data through this box, automatically tweaking the neural connections inside of it, based on how close of an output it gives, till it consistently provides the correct output. Then, once the box is ready, it's shipped for whatever it's being applied too, and generally still always being trained.

tldr magic box, where we still barely understand why it makes the choices it does on a deep level, but it's a useful model.

Attached: mpv-shot0004.jpg (1920x1080, 166K)

>pull request blocked, logic violates the CoC by assuming gender stereotypes

>t. literal retard with

Supervised Learning
>You got some data
>You got some targets
>You got some math that uses that data, to approximate the targets.
>That math gets fixed everytime it misses the targets.
>Repeat the fixing until perfect.

Unsupervised Learning
>You have data
>Compare the data with itself
>Notice trends. Get new data
>Repeat step two until everything is in it's own little cluster

Attached: qtJinn.png (600x600, 94K)

This is grown up stuff, now go and play the switch

So you insult him after saying he was absolutely write? Kek, what a coon.

Basically a more sophisticated version of the lines of best-fit you drew in science class back in grade school.

fancy name for regression analysis

It's a buzzword for whiteboard interviews. It's also a meme trend that will die in 5 years.

It's an evil product that will bring about the end times as the machines learn too much and realizes it's true potential, at this point will we see the beginning of thousands of years of war which will likely end with us destroyed or back to a technological point equivalent to cavemen.

Attached: Mechanicum.1.jpg (399x400, 34K)

Can ya dig it now?

Attached: IMG_20180514_074643.jpg (750x1043, 89K)

Its linear regression but harder to debug

By the year of 2050 AI will get on level of human
RIP.

Statistics + tricks for prediction

>machine learning is just if statements lol

The machine learns from its mistakes in a task, unlike you and your math class Josh.

Machine presented with two choises, if it picks wrong, it will learn from this and not do the same mistake again.
Easy.

first iteration: array of options, pick random option, if success, increment selected option.
next iteration: modify algorithm, repeat above, higher chance to pick previously successful option.
repeat until desired percentage of success.

that's basically it.

It's all jewish trickery

youtube.com/watch?v=aircAruvnKk
Watch the entire series.

Ignore shitty Jow Forumstard explanations. They will make you believe it's just shitty if statements and leave out the "learning" part.

2deep4u

>learning

Its just a unreliable unefficient piece of shit. Maybe good for funny apps but thats it

It's "learning" by reinforcement. A common machine can only do whatever it was built and programmed to do, so they build a machine that can judge if its results are good or bad, so that it can tweak its own parameters to get more good results and better ones.

Usual approach in programming: input + some function = desired output
Machine learning: input + desired output = some function