Start with matrix with 80% of the entries 0

In[361]:=

"svd-smooth_1.gif"

Out[364]=

"svd-smooth_2.gif"

Do SVD, set highest k eigenvalues to 0, reconstruct matrix

In[370]:=

"svd-smooth_3.gif"

Out[374]=

"svd-smooth_4.gif"
"svd-smooth_5.gif"
"svd-smooth_6.gif"
"svd-smooth_7.gif"
"svd-smooth_8.gif"
"svd-smooth_9.gif"
"svd-smooth_10.gif"
"svd-smooth_11.gif"
"svd-smooth_12.gif"

Bottom line : the more singular values you discard, the more smooth the matrix becomes


Created by Wolfram Mathematica 6.0  (30 January 2008) Valid XHTML 1.1!