SVD selection

[WWD]

In the AI and Machine Learning module, there exists a document in Lesson 2 “Data Preparation and Model Selection” title “Singular Value Decomposition”.  In the contained example, three SVDs are calculated.  The values of the SVDs are: 1.63, 0.49, and 1.45.  

The author of the document says that the SVDs are ordered by the magnitude of their associated SVD values.  In the example, SVD2 (value = 0.49) is said to be more important than SVD3 (value = 1.45).  That SVD2 is said to be more important than SVD3 is contrary to the method of assigning importance by the magnitude of the SVD.

My question is, why was SVD2 selected over SVD3, if magnitude of the SVD is used to select (or remove) a SVD?

Thank you,

Bill Donaldson

[mkeintz]

Could you provide a link to the source?

Also, is it possible that 0.49 is a typographical error?  If the value actually is 1.49, then SVD1>SVD2>SVD3.

[PaigeMiller]

In the contained example, three SVDs are calculated.

Are these the "eigenvalues" or "singular values"? Does the documentation use some other name for them where you have called them SVDs?

[Ksharp]

@Rick_SAS  wrote several blogs about SVD . you can answer in it.

https://blogs.sas.com/content/iml/2017/08/30/svd-low-rank-approximation.html

https://blogs.sas.com/content/iml/2017/08/28/singular-value-decomposition-svd-sas.html

https://blogs.sas.com/content/iml/2019/12/16/svd-de-noise-image.html

[Cynthia_sas]

Hi:
There are 5 classes in the AI&ML module. Which one of the 5 classes are you referring to? Are you asking about something in one of the video lectures? Asking about a demo? Asking about an exercise? Or, are you asking about a reference in the Course Notes. In order to an instructor to comment on your question, we need to know which class in the module you're asking about, since there are 5 classes and each class has a Lesson 2.
Thanks, in advance,
Cynthia