Support Vector Assertion - Practical Machine Learning Tutorial with Python p.22

In this tutorial, we cover the assertion for the calculation of a support vector within the Support Vector Machine.

[EatShitCh]

man you're amazing, thank you!
ps please never stop spreading knowledge

[zafarahmad]

A big thank you here. Whole series is not just very comprehensive but pellucid as well.

[chamberhorsefeeder]

I am not sure if it is mentioned during the video, but the projection of u to w is the dot product of u and the w's unit vector, not w.
u.w/||w|| = u*cos(alpha)= projection of u on w.

[neillunavat]

It is sad to see how the views go down with each video... Hope all reading this don't give up and be successful machine learning engineers.

[HilaryKansiime]

Love the fact that you explain the principles involved

[rajivkumar]

How to get the we get the output of u.w+b as -1 or 1? why not other numbers?Please explain.

[michaelwalczyk]

Wonderful video series! I'm learning a ton from it. I had a few questions related to this video.

At around 5:05, you introduce the formal notion of a "support vector" as X sub -sv and X sub +sv. What exactly do these variables denote? Are these the data points that are "closest" to the decision boundary (on either side)? Also, what happened to the inequality, i.e. <= 0 or >= 0? How do we know that the equations at 5:05 equate to -1 and +1?

Finally, I'm confused by Y sub i. Is this just a constant that we introduce for clarity? At around 8:30, why do we multiply the two equations by this value?

I know that's a lot...but any help would be appreciated. Thanks so much!

[thecactus]

I'm probably late but I get confused at 2:55. When you project u onto w you say u is in the + side of the hyperplane if u * w + b >= 0.

Couldnt you just check if u projected onto w > ||w|| or something?

[JohanSebastianCorn]

could someone explain why it has to be -1 or 1? like what determined this value? from 5:10 to 5:40

[bornofdata]

y(xw + b) - 1 = 0 is not the same for both, remember y is positive and negative for the other.

[becauseiwanttoanime]

so first of all what is like, what's the, we know that eventually xD nice one
thanks for an amazing video i couldn't understand the math so i had to go search on my own for a couple of days but still thanks for the tutorials

[huojinchowdhury]

Why we need to multiply class y with equation x.w +b?

[irishuserk]

Great video! Yourr really making machine learning accessible! I have a question on the value of b, could you explain its value a little more? When you say it is the bias, is that the spread between the two support vectors?

[mohammednagdy]

I don't get the part with X_sv and X+sv what are those?

[Joe-cszk]

I'm pretty confused as to why you're not using the projection formula to determine length to the decision boundary.

[李云龙-cq]

I really love your video.You explain things very well.Could you make some videos about other algorithms like random forest, decision tree:)

[RishikavsAnnie]

Could the magnitude on the RHS of these equations have been anything other than 1?

[johnshepard]

But why these values are 1 and -1? It is just an example and these can be 3 and 5 or anything else?

[chandeepadissanayake]

Did you ever explain what is a support vector? All the previous videos in this series helped me out in understanding from scratch(though I learnt the math behind linear regression from Khan Academy), this makes no sense to me.

[SanketPatole]

Isn't the projection of vector u on vector w = u.w/||w|| ?
Why did you consider only the dot product?