In mathematics, given a field , nonnegative integers , and a matrix , a rank decomposition or rank factorization of A is a factorization of A of the form A = CF, where and , where is the rank of .

Existence edit

Every finite-dimensional matrix has a rank decomposition: Let   be an   matrix whose column rank is  . Therefore, there are   linearly independent columns in  ; equivalently, the dimension of the column space of   is  . Let   be any basis for the column space of   and place them as column vectors to form the   matrix  . Therefore, every column vector of   is a linear combination of the columns of  . To be precise, if   is an   matrix with   as the  -th column, then

 

where  's are the scalar coefficients of   in terms of the basis  . This implies that  , where   is the  -th element of  .

Non-uniqueness edit

If   is a rank factorization, taking   and   gives another rank factorization for any invertible matrix   of compatible dimensions.

Conversely, if   are two rank factorizations of  , then there exists an invertible matrix   such that   and  .[1]

Construction edit

Rank factorization from reduced row echelon forms edit

In practice, we can construct one specific rank factorization as follows: we can compute  , the reduced row echelon form of  . Then   is obtained by removing from   all non-pivot columns (which can be determined by looking for columns in   which do not contain a pivot), and   is obtained by eliminating any all-zero rows of  .

Note: For a full-rank square matrix (i.e. when  ), this procedure will yield the trivial result   and   (the   identity matrix).

Example edit

Consider the matrix

 

  is in reduced echelon form.

Then   is obtained by removing the third column of  , the only one which is not a pivot column, and   by getting rid of the last row of zeroes from  , so

 

It is straightforward to check that

 

Proof edit

Let   be an   permutation matrix such that   in block partitioned form, where the columns of   are the   pivot columns of  . Every column of   is a linear combination of the columns of  , so there is a matrix   such that  , where the columns of   contain the coefficients of each of those linear combinations. So  ,   being the   identity matrix. We will show now that  .

Transforming   into its reduced row echelon form   amounts to left-multiplying by a matrix   which is a product of elementary matrices, so  , where  . We then can write  , which allows us to identify  , i.e. the nonzero   rows of the reduced echelon form, with the same permutation on the columns as we did for  . We thus have  , and since   is invertible this implies  , and the proof is complete.

Singular value decomposition edit

If   then one can also construct a full-rank factorization of   via a singular value decomposition

 

Since   is a full-column-rank matrix and   is a full-row-rank matrix, we can take   and  .

Consequences edit

rank(A) = rank(AT) edit

An immediate consequence of rank factorization is that the rank of   is equal to the rank of its transpose  . Since the columns of   are the rows of  , the column rank of   equals its row rank.[2]

Proof: To see why this is true, let us first define rank to mean column rank. Since  , it follows that  . From the definition of matrix multiplication, this means that each column of   is a linear combination of the columns of  . Therefore, the column space of   is contained within the column space of   and, hence,  .

Now,   is  , so there are   columns in   and, hence,  . This proves that  .

Now apply the result to   to obtain the reverse inequality: since  , we can write  . This proves  .

We have, therefore, proved   and  , so  .

Notes edit

  1. ^ Piziak, R.; Odell, P. L. (1 June 1999). "Full Rank Factorization of Matrices". Mathematics Magazine. 72 (3): 193. doi:10.2307/2690882. JSTOR 2690882.
  2. ^ Banerjee, Sudipto; Roy, Anindya (2014), Linear Algebra and Matrix Analysis for Statistics, Texts in Statistical Science (1st ed.), Chapman and Hall/CRC, ISBN 978-1420095388

References edit

  • Banerjee, Sudipto; Roy, Anindya (2014), Linear Algebra and Matrix Analysis for Statistics, Texts in Statistical Science (1st ed.), Chapman and Hall/CRC, ISBN 978-1420095388
  • Lay, David C. (2005), Linear Algebra and its Applications (3rd ed.), Addison Wesley, ISBN 978-0-201-70970-4
  • Golub, Gene H.; Van Loan, Charles F. (1996), Matrix Computations, Johns Hopkins Studies in Mathematical Sciences (3rd ed.), The Johns Hopkins University Press, ISBN 978-0-8018-5414-9
  • Stewart, Gilbert W. (1998), Matrix Algorithms. I. Basic Decompositions, SIAM, ISBN 978-0-89871-414-2
  • Piziak, R.; Odell, P. L. (1 June 1999). "Full Rank Factorization of Matrices". Mathematics Magazine. 72 (3): 193. doi:10.2307/2690882. JSTOR 2690882.