《机器学习Python实践》#1 课程介绍 – 译学馆

• 科普

SCIENCE

英语

ENGLISH

科技

TECHNOLOGY

MOVIE

FOOD

励志

INSPIRATIONS

社会

SOCIETY

TRAVEL

动物

ANIMALS

KIDS

卡通

CARTOON

计算机

COMPUTER

心理

PSYCHOLOGY

教育

EDUCATION

手工

HANDCRAFTS

趣闻

MYSTERIES

CAREER

GEEKS

时尚

FASHION

• 精品课
• 公开课
• 欢迎下载我们在各应用市场备受好评的APP

点击下载Android最新版本

点击下载iOS最新版本

扫码下载译学馆APP

《机器学习Python实践》#1 课程介绍

Practical Machine Learning Tutorial with Python Intro p.1

Well, girls and guys and welcome to an

in-depth machine learning tutorial series.

The objective of this tutorial series is to give you a holistic understanding of

machine learning and how it works and we’re going to be doing this by covering

a variety of algorithms so first we’re gonna be covering regression then we’re

going to be moving into classification with k-nearest neighbors and support

vector machines and then we’re going to get into clustering with flat clustering

a hierarchical clustering and then finally we’ll be getting into deep

learning with neural networks.

So, each step of the way with each of the major algorithms

we’re going to cover a theory, application and then we’re going to dive

in deep to the inner workings of each of them

so with theory this is the high-level intuitions it’s actually pretty quick to

digest the theory of all the algorithms. Most of the algorithms are actually

fairly basic since they need to be able to scale to very large amounts of data.

Then we move on to application this is also fairly quick

we’re going to use a module like scikit-learn and apply these algorithms

to real-world data to see exactly how they work what kind of input they expect

from us and what kind of output we should expect back from them.

Then finally we dive into the inner workings in the way that we’re going to do this

is by recreating the algorithms themselves from scratch in code

including all the math and all that so when you do this you’ll be able to have

a truly complete understanding of how these algorithms work which is going to

So, in order to fall along with a series i strongly

suggest you at least have the basics of Python 3 down understood if you don’t

I do have a Python 3 basics tutorial series here

check that out you really just need to get at least to the point where we

install modules with pip.

After that you can continue on but you really need those that initial I think

it’s like maybe 10 or 15 videos of core understanding of Python. We’ll also be

covering a healthy amount of math here but we’ll be talking about the matting

and explaining that as we go for sure so you’re not really expected to know much

becoming mostly algebra and geometry that comes up.

So, machine learning was really came about in the fifties so like more than a

half century ago now and it was defined in 1959 by Arthur Samuel as the field of

study where we give machines the ability to learn without being explicitly

programmed to do so.

So, the way I kind of like to think of it is it’s in viewing knowledge to a

machine without hard coding that knowledge so interestingly enough when I

talk to people both programmers and non-programmers and and and find out

what they know about machine learning

most people think machine learning is hard-coded so when you ask them what’s

wrong with machine learning or how is it different from actual learning that’s

where you find that most people think it’s hard coded so kind of interesting

issue there that most people are completely unaware that this field

exists and it’s actually not being hard coded so that was one machine learning

was actually kind of defined in 1963 Vladimir Vapnik came up with the support

vector machine but this really went pretty much overlooked until the

90s so Vladimir Vapnik was in the Soviet Union and then in the 90s he

was actually scooped out of the Soviet Union by bell labs and that’s when he

showed that that the support vector machine was better than the neural

network at the time doing handwritten character recognition so it’s

handwritten digits of our call right but anyway it was better than the support

vector a better than the neural network at that task the support vector machines

really took the lead for quite some time really up until very recently when

Google basically has kind of come back to really put some weight behind the

neural network specifically with deep learning.

But if you think you’re kind of late to the to the party

so to speak with machine learning I assure you

you’re not because I mean think about computers in the 95s i mean

we’re talking

we just started putting transistors on printed circuits instructions to your

computer was a maximum of a handful of bits at a time so pretty bad i mean even

this was very hard at

even if you were like a PhD student is very difficult to even get access to a

machine that could run significantly a support vector machine at scale [let’s say].

Whereas nowadays we live in a time where you can engage in deep learning with

neural networks on like gigabytes or even terabytes of data and what you can

do is you can spin up a you know hundred-thousand-dollar GPU cluster on

amazon web services and basically just rent it for a few dollars an hour and

then be done with it

that’s incredible like we living an incredible time this is the best time to

be alive is I think right now is the first time we’ve really been able to

really stretch and flex the muscles of machine learning up to this point it’s

really been learning without the machine part so we also so much to the point

where like with scikit-learn you can you can use scikit-learn with almost no

understanding at all you just apply it and you can usually get about

90~95% accuracy without messing with the default parameter

yeah you can just get it with the default parameters so that’s also pretty

crazy right

you want to push the limits and get more accuracy out of it then you need to

learn how they work and how you can tweak those parameters so you’re working

on self-driving car getting 90 to 95% accuracy and

identifying the difference between like a blob of tar and a child in a blanket

that’s not good enough you need much any more accuracy than that

so anyway that’s what this series is for is for the people are really looking to

push the limits on what’s available so if you really just want to learn the

basics actually already have some simple machine learning tutorials out there for

just applying machine learning to a dataset you can do this actually very

very fast.

So anyways the first topic that we’re gonna be covering is regression and

let’s go ahead and get into it.

[B]倔强