For example, We have a 4 * 4 Matrix a : [a, b, c, d] [e, f, g, h] [i, j, k, l] [m, n, o, p]]
Then, we can store an 2 * 2 block matrices to file. block(0, 0) is a 2 * 2 sub-matrix as below. [a, b] [e, f] Key : block(0, 0), Value : submatrix [a, b], [e, f] Key : block(0, 1), Value : submatrix [c, d], [g, h] Key : block(1, 0), Value : submatrix [i, j], [m, n] Key : block(1, 1), Value : submatrix [k, l], [o, p] ..... Also, If we can access sub-matrix from Hbase table, blocking can be pre-computed. On Tue, Oct 21, 2008 at 10:18 AM, Samuel Guo <[EMAIL PROTECTED]> wrote: > should the matrices that do the multiplication have the same blockID layout? > > And How to divide a matrix into BlockIDs to make the parallel computation > most efficiently? > > 2008/10/21 Edward J. Yoon <[EMAIL PROTECTED]> > >> Can anyone review this ?? >> >> http://blog.udanax.org/2008/10/parallel-matrix-multiply-on-hadoop.html >> >> -- >> Best regards, Edward J. Yoon >> [EMAIL PROTECTED] >> http://blog.udanax.org >> > -- Best regards, Edward J. Yoon [EMAIL PROTECTED] http://blog.udanax.org
