LFD Book Forum  

Go Back   LFD Book Forum > Course Discussions > Online LFD course > Homework 7

Reply
 
Thread Tools Display Modes
  #1  
Old 04-04-2013, 10:18 AM
udaykamath udaykamath is offline
Junior Member
 
Join Date: Jan 2013
Posts: 9
Default SVM equation from Slides

Prof Yaser.
Even though its not related to book chapter but the slide but thought it made sense here in this discussion.

In SVM slides Lecture 14

On page 13 we had just converted the problem to minimization and had

minimize Lagrangian(alpha)=....

On page 14 we have

maximize Lagrangian w.r.t alpha subject to...

How did this change?

Also page 15

we convert the maximize problem to minimize problem by inverting the sign.

So the jump from first minimize to maximize of lagrangian is not clear.

Thanks
Uday Kamath
Reply With Quote
  #2  
Old 04-04-2013, 01:12 PM
yaser's Avatar
yaser yaser is offline
Caltech
 
Join Date: Aug 2009
Location: Pasadena, California, USA
Posts: 1,477
Default Re: SVM equation from Slides

Quote:
Originally Posted by udaykamath View Post
In SVM slides Lecture 14

On page 13 we had just converted the problem to minimization and had

minimize Lagrangian(alpha)=....

On page 14 we have

maximize Lagrangian w.r.t alpha subject to...

How did this change?
Thank you for asking. In slide 13, minimization is w.r.t. some of the variables, and for the rest of the variables it is maximization (per Lagrange/KKT method). The minimization has already been done in the derivation in slide 13, so only the maximization part remains in slide 14.
__________________
Where everyone thinks alike, no one thinks very much
Reply With Quote
  #3  
Old 04-04-2013, 02:38 PM
udaykamath udaykamath is offline
Junior Member
 
Join Date: Jan 2013
Posts: 9
Default Re: SVM equation from Slides

thanks for your answer, so Lagrange(w,b,alpha) minimization means w.r.t to minimize w and b and maximize w.rt. alpha.

Also, you mention in the video that the KKT condition, the first one, of replacing the min yn(wXn+b) =1 is equivalent to using the inequality using the slack time square and adjusting. You mention that you will explain that in the Q&A, but no one asked that in Q&A and was wondering if you can here or some place give and explanation of the slack square thing and how min gets changed to non minimum with adding equality.

Thanks again for the wonderful lectures and book!
Forever indebted!
Uday
Reply With Quote
  #4  
Old 04-04-2013, 03:17 PM
yaser's Avatar
yaser yaser is offline
Caltech
 
Join Date: Aug 2009
Location: Pasadena, California, USA
Posts: 1,477
Default Re: SVM equation from Slides

Quote:
Originally Posted by udaykamath View Post
thanks for your answer, so Lagrange(w,b,alpha) minimization means w.r.t to minimize w and b and maximize w.rt. alpha.

Also, you mention in the video that the KKT condition, the first one, of replacing the min yn(wXn+b) =1 is equivalent to using the inequality using the slack time square and adjusting. You mention that you will explain that in the Q&A, but no one asked that in Q&A and was wondering if you can here or some place give and explanation of the slack square thing and how min gets changed to non minimum with adding equality.

Thanks again for the wonderful lectures and book!
Forever indebted!
Uday
The slack argument is probably available online in writeups about KKT. The basic idea is to add a squared variable s^2 to one side of an inequality to make it equality, and because it is squared, there are no resrictions on the value of s itself.
__________________
Where everyone thinks alike, no one thinks very much
Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 07:06 PM.


Powered by vBulletin® Version 3.8.3
Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
The contents of this forum are to be used ONLY by readers of the Learning From Data book by Yaser S. Abu-Mostafa, Malik Magdon-Ismail, and Hsuan-Tien Lin, and participants in the Learning From Data MOOC by Yaser S. Abu-Mostafa. No part of these contents is to be communicated or made accessible to ANY other person or entity.