how can I define k to ignore the small singular values?

how can I define k to ignore the small singular values?

References

Reference NO 1

http://www.mathworks.com/moler/eigs.pdf

Reference NO 2

http://www-user.tu-chemnitz.de/~potts/paper/litsing.pdf

Reference NO 3

http://www.math.nyu.edu/faculty/goodman/teaching/RPME/notes/Section3.pdf

Reference NO 4

http://stackoverflow.com/questions/17682626/singular-value-decomposition-different-results-with-jama-pcolt-and-numpy

Reference NO 5

http://docs.scipy.org/doc/numpy-dev/reference/generated/numpy.polyfit.html

Reference NO 6

http://www8.tfe.umu.se/courses/systemteknik/Media_signal_processing/04/presentations/MSP_P3-3.pdf

Reference NO 7

http://www.alterlab.org/teaching/BIOEN6670-1_2012/papers/Golub_1964.pdf

Reference NO 8

http://alias-i.com/lingpipe/demos/tutorial/svd/read-me.html