2012/12/14 Gael Varoquaux <[email protected]>:
> On Fri, Dec 14, 2012 at 05:51:08PM +0100, federico vaggi wrote:
>> In this case, X is:
>
>> (n_samples, n_features, M_bootstrapped_iterations):
>
>> So I thought I could get a better estimate by taking each M-length vector and
>> calculating the covariance against every other vector, which would result in 
>> a
>> (n_samples * n_features, n_samples* n_features) sized matrix, not a 
>> (n_samples,
>> n_features) sized matrix.
>
> Compute an (n_features, n_features) covariance matrix for each
> bootsrapped iteration and average them.

Also if the individual covariance matrix estimates are sparse, it
might be worth do a majority vote over the bootstrapped variants to
find a reasonably robust yet sparse support otherwise the averaging
itself might break the sparsity.

--
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to