about the approximation error

[mingminzhen]

    i have rerun the code and found that the error is different from the result described in your paper?

the error in the paper is computed ||k-k'||/||K||, but only the ||k-k'|| is computed in the code.

[mli]

hi mingmin,

the codes output ||k-k'|| because i often precompute and store the value ||k||, because it is a const value, and is not necessary to compute it every time.

best mu

On Sun, Dec 21, 2014 at 11:03 AM, mingminzhen [email protected] wrote:

i have rerun the code and found that the error is different from the result described in your paper?

the error in the paper is computed ||k-k'||/||K||, but only the ||k-k'|| is computed in the code.

— Reply to this email directly or view it on GitHub https://github.com/mli/nystrom/issues/1.