In this video I explain how SVM (Support Vector Machine) algorithm works to classify a linearly separable binary data set.
The original presentation is available at http://prezi.com/jdtqiauncqww/?utm_campaign=share&utm_medium=copy&rc=ex0share

Text Comments (419)
## Would you like to comment?

Akin O. (1 year ago)

Had to come back to revise my original comment.
This is the *best tutorial on SVMs* that I have ever come across (I have been through quite a few).
Really.

Thales Sehn Körting (10 months ago)

Thanks a lot.
Please subscribe to my channel.
Regards

Thales Sehn Körting (10 months ago)

Thanks a lot for your positive feedback.
Please like and share the video and subscribe to my channel.
Regards

rick schneider (1 year ago)

Ay Kaytee i had the same thought !

Elsio Antunes (15 days ago)

Que sotaque brazuca hein Tales?

Thales Sehn Körting (13 days ago)

100% Brasil :)
Um abraço e inscreva-se no canal

Simarjeet Gill (1 month ago)

Great representation... Much appreciated :3

Murilo Guimaraes (1 month ago)

Great explanation! You must be Brazilian by your accent right??

Thales Sehn Körting (1 month ago)

Certo!
Por favor inscreva-se no canal .
Um abraço

Ronit ganguly (2 months ago)

Best lullaby on SVMs

Thales Sehn Körting (2 months ago)

Subscribe to my channel and have a good night :)

Jacopo Solari (2 months ago)

should mention that SVM can be used for regression as well..

Tavvs Alves (2 months ago)

Excellent video.

Thales Sehn Körting (2 months ago)

Muito obrigado pelo comentário!
Um abraço

Henrique Bueno (3 months ago)

Your pronunciation is very good. Any non-native english can understand.

Thales Sehn Körting (2 months ago)

Thanks a lot for your comment. Please like/share/subscribe.
Regards
Talvez você seja brasileiro como eu? :)

Skyblade22 (3 months ago)

Incredibly painful to listen to, but the best explanation I've found. Audio quality could use a lot of improvement

Thales Sehn Körting (2 months ago)

Thanks for your comment. You are right about the bad sound quality. Unfortunately YouTube does not allow a simple fix in the video (e.g. boost the volume).
Regards

Zenville Erasmus (3 months ago)

Nicely done, sir!

Thales Sehn Körting (2 months ago)

Thanks for your comment.
Please like/share/subscribe.
Regards

supergo1108 (3 months ago)

Very clear! good video thanks!

Thales Sehn Körting (2 months ago)

Thanks for your comment here as well.
Please like/share/subscribe.
Regards

Aayush Poudel (4 months ago)

Excellent tutorial, Mr Thales Sehn Körting
! Thanks very much.

Thales Sehn Körting (2 months ago)

Thanks for your comment.
Please like/share/subscribe.
Regards

Zbynek Bazanowski (4 months ago)

You are presenting two versions of g(x). Neither of them yields the real distance of the points from the plane. The real distance is 1.12 and -1.12 for (2, 3) and (1, 1), respectively. The g(x) giving the real distance is x_1/sqrt(5) + x_2*2/sqrt(5) - 11/2/sqrt(5).

peyman morassai (4 months ago)

you are the best Thales, thank you so much...

Thales Sehn Körting (2 months ago)

Thanks for your comment.
Please like/share/subscribe.
Regards

dirtysock (5 months ago)

Really prime explanation, many thanks mane!

Thales Sehn Körting (2 months ago)

Thanks for your comment.
Please like/share/subscribe.
Regards

Ali (6 months ago)

But how does the algorithm works when you have more than 2 features? So for example your dataset consists of x1, x2, x3 and x4 and you want to predict Y, like the iris dataset for example.

umesh sai (6 months ago)

Hi Thales. Your videos are awesome! Could you also cover the Kernel Trick and Gaussian Kernels in another video. The way you explain with the visualizations is good and I think this would benefit many. THANK YOU!

Thales Sehn Körting (5 months ago)

Thanks for your feedback and suggestion. I have included in my videos wish list.
Please subscribe to my channel
Regards

Aridane Alamo (6 months ago)

Thanks! very clear and very understandable english for non-english viewers too.

Thales Sehn Körting (5 months ago)

Thanks for your feedback
Please subscribe to my channel
Regards

Huang Wade (6 months ago)

thanks

Thales Sehn Körting (6 months ago)

You are welcome. Subscribe to my channel.
Regards

Sarah Prado Medeiros (6 months ago)

VOCÊ ME SALVOU, OBRIGAADA!! thankss a lott

Thales Sehn Körting (6 months ago)

Muito obrigado pelo comentário.
Não esqueça de se inscrever no meu canal.
Abraço

arjun hegde (7 months ago)

In this example, we know the points (1,1) and (2,3).. .How will we know what are the closes points in real world problem?

Thales Sehn Körting (7 months ago)

Thanks for your feedback,
please like and share the video and subscribe to my channel.
This is a supervised algorithm so we have the information of all vectors to the 2 classes. With this information you compute the minimum distance.
Regards

Omar Abd (7 months ago)

thank you very much, nice explanation

Thales Sehn Körting (7 months ago)

Thanks for your feedback,
please like and share the video and subscribe to my channel.
Regards

hu jiawei (7 months ago)

sloppy explanation

Thales Sehn Körting (7 months ago)

For more sloppy videos please subscribe to my channel

Asadur Rahman (7 months ago)

Dear Sir, it was an excellent explanation. Thank you very very much................

Thales Sehn Körting (7 months ago)

Thanks for your feedback,
please like and share the video and subscribe to my channel.
Regards

Louis Boursier (7 months ago)

Thank you, it was very clear despite it can easily be made complex. Well done!

Thales Sehn Körting (7 months ago)

Thanks for your feedback,
please like and share the video and subscribe to my channel.
Regards

Conjugate Gradient (7 months ago)

Hello, I would like to use your examples in my paper. Is it okay to do so, and cite you? If so, do you have a paper or webpage I can reference? Thank you.

Thales Sehn Körting (7 months ago)

See previous comments to see the prezi link for this presentation. Feel free to use and you can cite the YouTube video or my research gate link where I have some articles using SVM.
Please subscribe to my channel
Regards

Snehal Jaipurkar (8 months ago)

Some basics cleared

Hari Sankara (9 months ago)

Simple and Best way to explain. Thx alot.

Thales Sehn Körting (8 months ago)

Many thanks for your positive feedback
Please like/share/subscribe.
Regards

Steven Pauly (9 months ago)

why a, 2a--> why the 2a ??

Weisi Zhan (9 months ago)

The example is so illustrating and finally makes it clear what 'support vector' means. Thanks for sharing :-)

Thales Sehn Körting (8 months ago)

Many thanks for your positive feedback
Please like/share/subscribe.
Regards

Elisa Warner (10 months ago)

How did he go from g(x) = (2/5)(x_1)+(4/5)(x_2)-(11/5) from g(x)=x_1 + 2x_2 - 5.5? He multiplied by (5/2), but why? Doesn't that affect the result?

Liu daniel (10 months ago)

the total margin formula is unclear，more details should be provided.

Joe Siu (10 months ago)

speak quieter please you're too loud

Thales Sehn Körting (10 months ago)

Buy better headphones, clean your ears and subscribe to my channel

Sebastian F. (10 months ago)

1:38 why higher margin z1 than z2 means it is better hyperplane?

mohit duklan (10 months ago)

1.25 speed is nice . normal is sooo funnyyyy :D

naouel ouafek (10 months ago)

the best of the best thank you !

Thales Sehn Körting (10 months ago)

Many thanks for this positive feedback. Please like and share the video and subscribe to my channel. Regards

Bala Kiswe (11 months ago)

watch video with 1.5X speed....

birinhos (11 months ago)

What is the "T" in g(x)= w^"T" .... ? Transpose ? Transformed ?

FrantzGray (10 months ago)

Transpose

Umair Hussain (11 months ago)

Thanks alot Sir!
love your accent btw :))

Thales Sehn Körting (11 months ago)

many thanks for your feedback. Please subscribe to my channel.
Regards

Hani YOUSFI (1 year ago)

Thank you so much.. So clear and simple.

Oliver Longhi (1 year ago)

Can you make a video to explain how to use svm to solve non linearly separable problems as well?

Thales Sehn Körting (10 months ago)

Thanks for the feedback.
Please like and share the video and subscribe to my channel.
I will put your suggestion in my wish list for next videos.
Regards

Talita Anthonio (1 year ago)

5a = 2, how did you get to that? Shouldn't it be -2?

Akin O. (1 year ago)

Excellent work sir!

Joan Perez Guallar (1 year ago)

I was starting to drift off, dude speak a bit faster

Thales Sehn Körting (10 months ago)

Some comments already suggest to use 1.5x speed :)

Mark Misin (1 year ago)

Thank you for the video! Very well done! I am a bit confused about how you went from g(x)=2/5*x1+4/5*x2-11/5 to g(x)=x1+2*x2-5.5 in the end. You can only get that form if you multiply both sides of the equation by 5/2, but then it would be 5/2*g(x)=x1+2*x2-5.5

Canal doFuba (1 year ago)

maravilha

ultra (1 year ago)

@6:27 "support vectors" are defined as (2/5, 4/5). I was always under the impression that the 'support vectors' are the ones used to make the hyperplane (i.e. (1,1) and (2,3). Please advise.

Amimul Ummah Baiqi (1 year ago)

Dear Mr. Thales Sehn Korting
This video very usefull to me to do my thesis, but I want to ask you about the value of z1 and z2, why you said in this video Z2 is higher then Z1?
I will waiting your respon,
Thankyou sir

Amimul Ummah Baiqi (1 year ago)

Dear Mr thales, Can I get your contact,? Because I want to learn about SVM multiclass. I am using that to solve my thesis.

Thales Sehn Körting (1 year ago)

Thanks for your feedback and question.
Please like and share this video and subscribe to my channel.
Around 1:13 I try to show visualy that z2 is higher than z1. I mean, the idea is that the distance between the two classes in the example must be maximized. And SVM tries to find exactly the hiperplane with the maximum distance. In the case of the video, z2.
Regards

Deep Net (1 year ago)

Thank you. Great Video.

Thales Sehn Körting (1 year ago)

Thanks for your feedback, please watch my new video about CNN.
Please like and share this video and subscribe to my channel.
Regards

johnfy.k hikc (1 year ago)

Hey, there's one of these about LOGISTIC REGRESSION?

Thales Sehn Körting (1 year ago)

Thanks for your suggestion, I will put on my wish list for next videos.
Please like and share this video and subscribe to my channel.
Regards

Mr Schwszlsky (1 year ago)

But how you sure us that realy work in much data??

Mr Schwszlsky (1 year ago)

can you tell me the best size of training sample in using SVM??

Thales Sehn Körting (1 year ago)

Thanks for your question. The notion of much data is pretty much relative. We also have the way that the algorithms has been developed, the capacity of the computer to store several data in memory, etc. After this, if the clusters behave in the way we described on the video, there is a good chance to work with "much data".
Please like and share this video and subscribe to my channel.
Regards

ndiayej100 (1 year ago)

Hands down THE best SVM explanation I have ever seen!!! Thank you, Sir!

Thales Sehn Körting (1 year ago)

Many thanks for this positive feedback!
Please like and share this video and subscribe to my channel.
Regards

zé (1 year ago)

I got you are brazillian (like me) in the first minute... Thanks for sharing your knowledge!

Thales Sehn Körting (1 year ago)

+zé sure I am Brazilian ;)
Se possível dê like e se inscreva no canal.
Um abraço

Ashish Jha (1 year ago)

Best video on SVM!

Thales Sehn Körting (1 year ago)

+Ashish Jha thanks for the positive feedback.
Please like and share the video and subscribe to my channel.
Regards

Muhammad Kamran Javed (1 year ago)

Is it the optimal hyperplane?

Kamal Sehairi (1 year ago)

Thank you Sir,
Simple and straightforward
If you can't explain it simply, you don't understand it well enough
I think you understand it very well and you presented it in very simple way

Thales Sehn Körting (1 year ago)

Many thanks for your feedback. I agree with you that such topics must be explained as simples as possible. Please like and share the video and subscribe to my channel.
Best regards

zamine81 (1 year ago)

thanks for the explanations, i would to know what is the highest number of classes or labels can be resolved by or used in SVM ? because in every introduction i saw only binary classification (two classes).

zamine81 (1 year ago)

@Thales Sehn Körting another question please, from what i have seen in documentation the training set elements ( called vectors if i'm not mistaken) are represented in two dimensional plan, but in the mathematical description a x element from training set is from R-d ( d dimensions ), so is this two dimensional representation is valid only for d=2? and in the case of d>=3 can be the x elements and the hyperplan separator represented geometrically ?

zamine81 (1 year ago)

@Thales Sehn Körting thank you a lot for this usefull explanation

Thales Sehn Körting (1 year ago)

+zamine81 thanks for your feedback. Please like and share the video and subscribe to my channel.
The SVM algorithm was designed to treat binary problems however it can be extended to a multiple number of classes, therefore there is not a limit for the number of classes.
Regards

Aziz Saouli (1 year ago)

Anyone have svm code in matlab ?????

Thales Sehn Körting (1 year ago)

Thanks for your question, please like and share the video and subscribe to my channel.
You should look for libsvm, which is probably available in several platforms.
Regards

mustafa muhamad (1 year ago)

What are the differences between SVM method and Naive Bayes method.

BEING SPIRITUAL (1 year ago)

Speed 4x needed

Thales Sehn Körting (1 year ago)

+RAJAT JAIN thanks for your feedback
Want more slow videos? Subscribe to my YouTube channel.
Regards

D Alexander (1 year ago)

Did not understand much.Could have made bit more simpler. Where are the equations coming from?

Thales Sehn Körting (1 year ago)

Thanks for your question. Please subscribe to my channel.
Check the references in the end of the video, from there I took only the most important equations, in order to simplify the video. I assure you will find much more equations in the references ;)
Regards

jansi rani (1 year ago)

you elucidate very well i understood SVM concept very clear.. Thank you so much

jansi rani (1 year ago)

:)

Thales Sehn Körting (1 year ago)

+jansi rani thanks for this positive feedback. Please like/share the video and subscribe to my channel.
Regards

Loop loop (1 year ago)

Do I need to learn advance linear algebra 1st ?

Thales Sehn Körting (1 year ago)

thanks for your feedback.
Indeed there is a lot of packages with SVM available (libsvm, R, Python, etc), so you don't need to know most of the details of the algorithm. However in this video I try to explain how it works, so that users will not use the algorithm as a black box.
regards, please subscribe to my channel

Zeyd Boukhers (1 year ago)

It is a perfect explanation.

Thales Sehn Körting (1 year ago)

many thanks for the feedback
please like/share the video and subscribe to my channel
regards

Quintin Lohuizen (1 year ago)

You really gotta speak louder... My ears aren't that good and the volume of my laptop is already on its max....

Thales Sehn Körting (1 year ago)

thanks for the feedback. you are right, this video was published with low volume. sorry for that.
in my next videos, the audio is better
so please like/share the video and subscribe to my channel
regards

RAMASUBRAMANIAN RAVI (1 year ago)

Useful vedio thank u

Thales Sehn Körting (1 year ago)

+RAMASUBRAMANIAN RAVI thanks a lot for your feedback. please like and share the video and subscribe to my channel. thanks regards

Biranchi Narayan Nayak (1 year ago)

This is the best video tutorial on SVM.

Thales Sehn Körting (1 year ago)

+Biranchi Narayan Nayak many thanks for your feedback. Please like and share the video and subscribe to my channel. Regards

祝晓 (1 year ago)

看不懂

Tsunami! :o (1 year ago)

Could you name your variables please? The vector w at 3:05 is being multiplied (dot prosuct) with the x vector. Why is w a vector and not a scalar?
I'm a confused boï ya'know!

Boris Dessimond (1 year ago)

Very clear and quick. Thanks!

Thales Sehn Körting (1 year ago)

+Boris D. Thanks for the feedback. Please like and share the video and subscribe to my channel. Regards

ForKSapien (1 year ago)

Best tutorial I've seen. Helps me a lot. Thank you very much!!

Thales Sehn Körting (1 year ago)

+ForKSapien thanks
please like/ share/ subscribe!
regards

sundar raman s (1 year ago)

Excellent Demo on SVM basics

Thales Sehn Körting (1 year ago)

+sundar raman s many Thanks for your feedback. please subscribe to my Channel. Regards

asdalow (1 year ago)

3:53 But I don't want to separate the chocolates, I want 'em all.

asdalow (1 year ago)

As he speaks he almost gets drown.
And as I listen to him and hearing him drowning, I feel that neither can I breathe .

asdalow (1 year ago)

Sorry about that :D
In the video I almost never hear you breathing. I think this is what made me feel like I'm drowning.
However, great video and thank you for the value you gave me with that! :)

Thales Sehn Körting (1 year ago)

+electroGadgets glub glub glub

Felipe Coutinho (1 year ago)

You are good at this. Keep making videos, your videos are the best.

Thales Sehn Körting (1 year ago)

+Felipe Coutinho thanks for the positive feedback.
Please like and share the video and subscribe to my channel.
Regards

jor4288 (1 year ago)

This video is very helpful.

Thales Sehn Körting (1 year ago)

+jor4288 thanks for the positive feedback.
Please like and share the video and subscribe to my channel.
Regards

Ulil Latifah (1 year ago)

wow

Thales Sehn Körting (1 year ago)

please like/share the video and subscribe
regards

Douglas Silva (1 year ago)

You are brazilian, aren't you?

Douglas Silva (1 year ago)

Opa thales! boa meu velho!

Thales Sehn Körting (1 year ago)

+Douglas Henrique sim \o/
não esqueça de se inscrever no canal
abraço

foo bar 167 (1 year ago)

w = ( (2,3) + (1,1) ) / 2 = (1.5, 2) -- is the middle point between two points (2,3) and (1,1)
You could check this fact if you're take a ruler and measure coordinates of w manually.
Also I don't understand where constant "a" is came from (4:38 - "...so we have a constant 'a'..." - this is not an explanation).
And what is the reason to multiply (2,3) on (a,2a) => 2a+6a+w=1? I don't understand the logic behind this multiplication.

lemus Aly (1 year ago)

q oso

Thales Sehn Körting (1 year ago)

+Alicia Lemus oso? :)

Keet Malin Sugathadasa (1 year ago)

Thank you very much sir

Thales Sehn Körting (1 year ago)

+Keet Malin Sugathadasa thanks for your feedback. Please like and share the video and subscribe to my channel. Regards

Dipti Borade (1 year ago)

Excellent Explanation! Thanks a lot!

Dipti Borade (1 year ago)

Yes I subscribed your channel! Sir, can you please make a video on multiclass SVM (one-vs-rest strategy)?

Thales Sehn Körting (1 year ago)

+Dipti Borade thanks a lot for your positive feedback
please like and share the video and subscribe to my Channel
best regards

김대한 (1 year ago)

Awesome!

Thales Sehn Körting (1 year ago)

+Peter Kim thanks for your feedback
please like and share the video and subscribe to my channel
regards

Siva prasanth (1 year ago)

Sir you are the best.. Even a student with no knowledge of machine learning can understand this complex machine learning algorithm if he sees this video.

Thales Sehn Körting (1 year ago)

thanks for your good feedback
please like/share the video and subscribe to my channel
regards

Michał Dobrzański (1 year ago)

deez distance ftw!

Kuka Tech (1 year ago)

Thales, I liked the tutorial, but I think you would explain that better in portuguese. Thank you anyway!

5iLikePie5 (1 year ago)

I think there is a mistake (not sure though): when you are getting the weight vector (a,2a), isnt it supposed to be a perpendicular vector to the one you get? The one u get(a,2a) is in direction of the line that goes through the 2 points and the line that separates (I guess the line that the weight vector is supposed to be on) is perpendicular to this one.
i.e. should the weight vector be -[a, 0.5*a] ? (derived from simple math like: y = mx , normal to this would be y = -mx)

Satyo Wicaksana (1 year ago)

what do we do when there is more than 1 support vector for each class?

Jugs Ma马家杰 (1 year ago)

wow, the explanation in this video made me really clear about SVM.

Thales Sehn Körting (1 year ago)

thanks for the feedback
please subscribe to my channel
regards

Douglas Monteiro (1 year ago)

are you Brazilian? you accent sounds a lot like the Brazilian one...

Saeed Nusri (1 year ago)

Great video! Thanks for the explanation.

Thales Sehn Körting (1 year ago)

thanks for the good feedback
please like/share the video and subscribe
regards

Tob Ias (2 years ago)

Thanks, I wonder what you do when the two classes overlap.

Just Me (2 years ago)

what horrible voice...I had a headache

Thales Sehn Körting (2 years ago)

thanks for your feedback
if you have a beautiful voice, please create a good video on SVM, then I will redirect this video to yours
regards

Wassauf Khalid (2 years ago)

ultra slow speed

Thales Sehn Körting (2 years ago)

How about a useful (and fast) comment? :)
Regards

Sefa Saylan (2 years ago)

Thanks for this video, it explains this topic very well

Thales Sehn Körting (2 years ago)

thanks for your feedback
please like/share the video and subscribe to my channel
regards

deependra singh jhala (2 years ago)

awesome :)

deependra singh jhala (2 years ago)

Thales Sehn Körting :) subscribed & shared

Thales Sehn Körting (2 years ago)

thanks for your feedback
please subscribe to my channel and share this video with your peers
regards

Đức Huy Lê (2 years ago)

in the lecture you use an equation to find the hyperplane but in the example you use geometry to solve. i'm so confused about the equation.

Cristian Alejandro Rojas (2 years ago)

Great explanation

Thales Sehn Körting (2 years ago)

thanks for your feedback
please like/share the video and subscribe.
regards

Ryan Keegan (2 years ago)

Great video thank you. Can you tell me how you got your final g(x) = x1 +2x2 -5.5 ? I understood everything up to how you got the last line.

guruphiji (2 years ago)

I assume you van multiply the whole equation by 5/2 because you are at the border (where g(X)=0). But how it applies to other points where we want |g(X)|>1 ?

Thales Sehn Körting (2 years ago)

Thanks for your question.
Please like/share the video and subscribe.
To get the final g(x), when we discovered the vector w = (2/5, 4/5) and w0, we used them in the original equation (see 1:50):
g(x) = w x + w0
Regards

akash aggarwal (2 years ago)

Very nice explanation sir!!!

Thales Sehn Körting (2 years ago)

thanks for your feedback
please like/share and subscribe to my channel
regards

Join YouTube for a free account, or sign in if you are already a member.

© 2019 Drunk and sleeping girls video

The Headache Paradox: It is scientifically impossible to have a headache, seeing that there are no pain receptors in the brain. Yet we continue to search for answers to what causes a headache, because we all know that headaches do exist. Between a combination of the fact that our brains are roughly nothing more than a highly advanced mass of nerves, and that every function is controlled by signals that our brain sends through our nerves to the rest of our bodies, I think that if we could somehow make our nerves send stronger, more efficient signals at a quicker speed it would have a lot of different results. Including heightened sense (both physical and psychic) moving quicker both running and walking along with arm movement etc. The brain being just nerves would obviously also be effected, resulting in quicker more efficient thinking, which naturally would make us smarter, more observant, and have much sharper reflexes. Some say that is an overly simplified form of how the brain functions, which it is. To them I say make the axons 12-15% wider, resulting in signals being sent much more quickly through the brain. Allowing neurons to create more connections without loss of speed, and the 10,000+ miles of blood vessels in the brain doubling as an additional cooling system for the brain. I could write a book on the cascade of effects that alone causes for people when they are changed in to a real vampire, however for now I will leave it in this simplified form.