Understanding Low Rank Approximation with SVD: A Comprehensive Guide

In summary, the conversation discusses the concept of low-rank approximation through SVD. The result matrix is expected to have a lower rank, but examples show higher ranks due to rounding errors. The process involves calculating the SVD and working with it in double precision arithmetic. Further steps in the feature extraction process may involve determining the basis and mapping terms to the basis. SVD decomposition is a useful tool in reducing large models to manageable sizes. In practice, only a subset of singular values may be computed to save time and resources.
  • #1
TheOldHag
44
3
I'm studying low rank approximation by way of SVD and I'm having trouble understanding how the result matrix has lower rank. For instance, in the link the calculation performed on page 11 resulting in the so-called low rank approximation on page 12. This matrix on page 12 doesn't appear to me to be of rank 2 as one would think. I would expect only 2 linearly independent row vector. When I put it back into a calculator it is still rank 9.

http://lsa.colorado.edu/papers/dp1.LSAintro.pdf
 
Physics news on Phys.org
  • #2
This is hard for me to word.

Basically, this explanation of low-rank approximation says the new m x n matrix should have rank r is not something I'm seeing in the Latent Semantic Indexing examples out there. They are coming up with new matrices that to me look like they have higher rank than r.

http://en.wikipedia.org/wiki/Low-rank_approximation
 
  • #3
The matrix on page 12 is only printed to 2 significant figures, so if you copied and pasted those numbers into a calculator, most likely the matrix wound be "rank 9" because of the rounding errors introduced.

The matrix on page is actually the product W (12 x 2) . S (2 x 2) . P' (2 x 9) where the ( ) are the dimensions of each matrix and the relevant parts of the full W S and P matrices are shaded on page 11.

Of course the full W and P matrices on page 11 are only shown to two significant figures as well. If you check whether W and P are orthonormal using those rounded values, you will probably find the "zero" terms are of order 0.0001 or worse.

If you want to do a numerical check for orthogonality, you really need to calculate the SVD yourself to 16 significant figures in double precision arithmetic, and work with that.

FWIW it should be obvious that if you took only one SVD component, the resulting matrix
W (12 x 1) . S (1 x 1) . P' (1 x 9)
would be of rank 1.
 
  • #4
I think your last point helps understand this better and the rounding error makes perfect sense.

As far as feature extraction in general or latent semantic indexing in general I'm not sure how this reduced dimensionality results in reduced computational cost down the road since you are still left with a term-document matrix that is no longer sparse but having fewer dimensions. Are there further steps to this feature extraction process beyond this point. For instance, determine the basis and then maps things in term of the basis. Seems to be a step missing in these papers on reading since they just end with this new matrix that doesn't appear much simpler on the surface.
 
  • #5
I don't know anything about semantic feature extraction specifically, but the math in your link looks quite similar to methods that are used in other areas to reduce "large" models to manageable size. Numerical methods based on the SVD decomposition don't seem to feature much in first courses in numerical methods, but they are very useful in practice.

I assume those examples show the matrices in full just to demonstrate what is going on. In practice, it is unlikely you would compute the full matrices. For example if you only want to compute a subset of the singular values (say 1000 SVs of a matrix of dimension maybe 10,000,000 x 20,000, for a large text data base with a realistic sized vocabulary), you don't need to calculate all the SVs , plus the full orthonormal matrices, and then throw most of that data away.
 
  • #6
I appreciate your help. I'm pass that bit of confusion now.
 

Related to Understanding Low Rank Approximation with SVD: A Comprehensive Guide

1. What is low rank approximation?

Low rank approximation is a mathematical technique used to approximate a matrix with a lower rank matrix. It is commonly used in data compression, signal processing, and machine learning to reduce the size and complexity of a matrix while preserving its important features.

2. How does low rank approximation work?

Low rank approximation works by decomposing a matrix into two or more smaller matrices, typically using techniques such as singular value decomposition or principal component analysis. These smaller matrices have a lower rank and can be used to approximate the original matrix with a certain level of accuracy.

3. What are the advantages of using low rank approximation?

Low rank approximation can significantly reduce the size of a matrix, making it more computationally efficient to work with. It also allows for easier interpretation of data by identifying the most important features and reducing noise. This can be particularly useful in machine learning and data analysis tasks.

4. What are the limitations of low rank approximation?

One limitation of low rank approximation is that it may not be suitable for all types of data. It is most effective for matrices with a clear underlying structure, and may not work well for highly complex or sparse data. Additionally, the level of approximation may not always be accurate enough for certain applications.

5. How is low rank approximation used in real-world applications?

Low rank approximation has a wide range of applications in various fields, such as image and video compression, data mining, and recommender systems. It is also commonly used in linear algebra and numerical analysis for solving large systems of linear equations. In machine learning, low rank approximation can help with feature selection and reducing overfitting in models.

Similar threads

  • Linear and Abstract Algebra
Replies
8
Views
3K
  • Linear and Abstract Algebra
Replies
5
Views
1K
  • Linear and Abstract Algebra
Replies
1
Views
1K
  • Linear and Abstract Algebra
Replies
5
Views
22K
  • Linear and Abstract Algebra
Replies
4
Views
1K
  • Linear and Abstract Algebra
Replies
1
Views
4K
  • Linear and Abstract Algebra
Replies
4
Views
2K
  • Linear and Abstract Algebra
Replies
1
Views
2K
  • Linear and Abstract Algebra
Replies
4
Views
2K
  • Special and General Relativity
Replies
6
Views
3K
Back
Top