mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Dunning <ted.dunn...@gmail.com>
Subject Re: SSVD Wrong Singular Vectors
Date Tue, 04 Sep 2012 16:27:11 GMT
A quick t-test on these differences gives the same results.... no
significant difference.

On Mon, Sep 3, 2012 at 11:34 PM, Dmitriy Lyubimov <dlieu.7@gmail.com> wrote:

> Then i subtracted error means between two methods (+ sign means
> smaller error for MR version, -sign means smaller error for R
> sequential version):
>
> -2.490580e-06  6.065328e-03  1.954353e-03  1.557188e-03  3.680165e-03
>  9.346332e-04 -6.052074e-04  2.622627e-03 -6.930656e-04  2.608697e-03
> -1.218882e-03  1.042435e-03  1.473642e-03  8.322338e-03  6.035364e-03
> -6.508083e-03  6.390501e-03 -8.441491e-03  1.169098e-02 -5.618290e-04
> -1.061915e-02 -7.908856e-03  2.295738e-03  2.397220e-04 -9.223938e-03
> -3.254358e-03  4.996947e-03 -5.478588e-04 -4.909232e-03 -3.874268e-03
>
> 95% confidence interval for the error mean measurements of MR version is,
> +/-:
>
> 5.754773e-06 9.002358e-03 9.890168e-03 1.057098e-02 7.995073e-03
> 9.310558e-03
> 1.002221e-02 5.687491e-03 7.069399e-03 8.315481e-03 9.091700e-03
> 1.286121e-02
> 1.265919e-02 1.859940e-02 1.523656e-02 1.398035e-02 9.846969e-03
> 3.852648e-02
> 1.943268e-02 2.668289e-02 2.574338e-02 2.376454e-02 1.330533e-02
> 1.425717e-02
> 2.280331e-02 2.086706e-02 2.326773e-02 2.981179e-02 2.949766e-02
> 2.059868e-02
>
> which seems to be consistently greater than difference in the errors.
> Also, we see intermittent pattern of one method taking precision edge
> over another.
>
> This basically tells me there's no statistically difference in
> precision between sequential and MR version errors with the same
> parameters.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message