@Article{Berry:2005:ACS,
  author =       "Michael W. Berry and Shakhina A. Pulatova and G. W. Stewart",
  title =        "Algorithm 844: Computing Sparse Reduced-Rank Approximations to Sparse
                 Matrices",
  journal =      "{ACM} Transactions on Mathematical Software",
  volume =       "31",
  number =       "2",
  month =        jun,
  year =         "2005",
  pages =        "252--269",
  URL =          " http://doi.acm.org/10.1145/1067967.1067972",
  abstract =     "In many applications --- latent semantic indexing, for example --- it is required 
                 to obtain a reduced rank approximation to a sparse matrix $A$.  Unfortunately, the 
                 approximations based on traditional decompositions, like the singular value and QR
                 decompositions, are not in general sparse.  Stewart [{\it Numer. Math.} 83 (1999) 
                 313--323] has shown how to use a variant of the classical Gram--Schmidt algorithm, 
                 called the quasi--Gram-Schmidt--algorithm, to obtain two kinds of low-rank
                 approximations.  The first, the SPQR, approximation, is a pivoted, Q-less QR 
                 approximation of the form $(XR_{11}^{-1})(R_{11} R_{12})$, where $X$ 
                 consists of columns of A.  The second, the SCR approximation, is of the form the
                 form $A \cong XTY^{\rm T}$, where $X$ and $Y$ consist of columns and rows
                 $A$ and $T$ is small.  In this paper we treat the computational
                 details of these algorithms and describe a Matlab implementations.",
}