LFD Book Forum  

Go Back   LFD Book Forum > Course Discussions > Online LFD course > Homework 6

Reply
 
Thread Tools Display Modes
  #1  
Old 05-13-2013, 07:07 PM
jlaurentum jlaurentum is offline
Member
 
Join Date: Apr 2013
Location: Venezuela
Posts: 41
Default What about residual analysis in linear regression?

I've been kind of saving this question, but decided to ask at this point.

Why is there no mention of residual analysis in any of the linear regression topics the course has covered? How does residual analysis fit into the data learning picture (if it fits in at all)?

Specifically: starting with this week's topic of regularization, we've seen how weight decay softens the weights, but in doing so, chages them from the normal weights you'd obtain in linear regression. I would imagine that with weight decay, it would no longer hold that the mean of the errors (as in linear regression errors: \hat{y}-y) is equal to zero, so the residuals would not be normally distributed with same variance and zero mean. In other words, with weight decay at least one of the Gauss-Markov assumptions do not hold?

Does that matter?

In general, are the standard tools of linear regression analysis we were taught in school (looking at the determination coefficient, hypothesis testing on the significance of the coefficients, and residual analysis to see if the assumptions that back up the previous elements hold) entirely pointless when you're doing machine learning?
Reply With Quote
  #2  
Old 05-13-2013, 07:38 PM
yaser's Avatar
yaser yaser is offline
Caltech
 
Join Date: Aug 2009
Location: Pasadena, California, USA
Posts: 1,474
Default Re: What about residual analysis in linear regression?

Quote:
Originally Posted by jlaurentum View Post
I've been kind of saving this question, but decided to ask at this point.

Why is there no mention of residual analysis in any of the linear regression topics the course has covered? How does residual analysis fit into the data learning picture (if it fits in at all)?

Specifically: starting with this week's topic of regularization, we've seen how weight decay softens the weights, but in doing so, chages them from the normal weights you'd obtain in linear regression. I would imagine that with weight decay, it would no longer hold that the mean of the errors (as in linear regression errors: \hat{y}-y) is equal to zero, so the residuals would not be normally distributed with same variance and zero mean. In other words, with weight decay at least one of the Gauss-Markov assumptions do not hold?

Does that matter?

In general, are the standard tools of linear regression analysis we were taught in school (looking at the determination coefficient, hypothesis testing on the significance of the coefficients, and residual analysis to see if the assumptions that back up the previous elements hold) entirely pointless when you're doing machine learning?
Residual analysis and other details of linear regression are worthy topics. They are regularly covered in statistics, but often not covered in machine learning. If you recall in Lecture 1, we alluded quickly to the contrast between statistics and machine learning (which do have a substantive overlap) in terms of mathematical assumptions and level of detailed analysis. Linear regression is a case in point for that contrast.
__________________
Where everyone thinks alike, no one thinks very much
Reply With Quote
  #3  
Old 05-14-2013, 06:31 AM
jlaurentum jlaurentum is offline
Member
 
Join Date: Apr 2013
Location: Venezuela
Posts: 41
Default Re: What about residual analysis in linear regression?

Thank you for the quick reply, Professor. I'll review lecture one more closely.
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 05:16 PM.


Powered by vBulletin® Version 3.8.3
Copyright ©2000 - 2018, 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.