Support Vector Machine Intro and Application - Practical Machine Learning Tutorial with Python p.20 Video Lecture | Machine Learning with Python - AI & ML

what is going on everybody who welcome to another machine learning

with height onto tutorial and this you toilet

you're talking about another supervised machine learning class a fire and

that is these support vector machine

or as the end

right out at the gate you should think about what is super vector machine what is it's even meanwhile first of all machine probably just has some sort of

[COUGH] unto like a system or something like

but then we have support what vector so

you should tell us that we're going to be dealing with vectors and in better space and then we see we have

port and we're not really sure what the heck of that supposed to mean but generally we're going to pride guess that it because together so we've got support vectors

not really sure with those are yet but what we figured out so we got

port vector machine we are looking prize for some support vectors and we are recognizing that we're working on vectors states

[UH] so the support vector machine was created by a lot of your nick back in the sixties

actually between largely ignored never looked until

ninety as the mic was in the us all at the time and then in the nineties you got moved over to the united states

working with bell labs and in the ninety this is where it was shown that the sport like a machine was better than the neural network at

i can written number recognition

this stuff like this and they basically became the

the most popular machine learning algorithm for a while and it still one of the

most popular machine learning algorithms so first let's just talk about the really high level intuition

no of the support back to machine and the most shown actual example of us using it and then i will break it down to

explained how it does when it does it turns out this protect machine i would

would you was probably one of the most complex machine learning algorithms so again working

vector space so a straw some vector space and then

the part of actor machine is what's called a binary classify so it's separates only

the two groups at a time it's important to not confuse this that a support vector machines can only classified

into two groups it just well separate two groups are basically separate groups

to really one at a time one from the rest i'll explain that more when we dive in but for now just understand it it's a by nearly class fire and what it wants to do is separate into two groups and these two groups are

and no to as positive or negative and generally they're going to be a positive one or a negative one so it's say you got two groups of data so you've got some positive so draw positive positive

positive and then you're going to have some negatives so i'll draw negative negative and a negative

and the objective of the sun is to find the best

operating hyper plane which is also referred to as your decision bow

~~~~~~~ so the best separating hyper plane decision boundary that will separate this data

so for now it's going to look a lot like a line and i think that's why a lot of people

can get confused because the spans are always depicted are usually depicted in two dimensional space at most three dimensional space

so the some is going to go it's going to run and it's going to say i found the best severing hyper plane and it is

it's like this when it would be a straight line if i could draw right so is thing that's the best supreme separating hyper plane and it's going to see that that is the best because the distance

looking that hyper plane and the the associated data that it's separating is the greatest so if you were to take a perpendicular by sector of the hyper plane and draw that to the closest data points that say to have

with this and this

that that distance

is the greatest distance that you could come up with with a separating however paint so for example another separating hyper plane

might be this right that's a separating hyper point but the distance again the perpendicular distance between those is this

it

this much smaller than our green as shallow

so this project machine says that yellow greenish line is the best separating hyper plane for our data now how we get to that is another story but for now let's just remember that is the let's say that's the best separating hyper plane [UM]

once you acquire the best separating hyper plane you can now take an unknown data so let's say you haven't

unknown data points so unknown you because that you're rests on the right hand side of that separating hyper plane or goes say it's not

and it's a positive sample converse

so if you had unknown that was down here

it's not a known it's on the left inside of this hike airplane

therefore it's actually a negative sample

so that is the intuition of these support vector machine