On Mon, Jun 29, 2009 at 10:22 PM, Robert Haas wrote:
> I'm finding myself unable to follow all the terminology on this thead.
> What's dimension reduction? What's PCA?
[snip]
Imagine you have a dataset with two variables, say height in inches
and age in years. For tue purpose of discussion lets
> I'm finding myself unable to follow all the terminology on this thead.
> What's dimension reduction? What's PCA?
( Sorry for the jargon - thanks Josh )
> It feels like what you might need is statistics for colB (MCVs and/or a
> histogram) for certain particular values of colA.
Certainly - th
On Mon, Jun 29, 2009 at 10:22:15PM -0400, Robert Haas wrote:
> I'm finding myself unable to follow all the terminology on this thead.
> What's dimension reduction?
For instance, ask a bunch of people a bunch of survey questions, in hopes of
predicting some value (for instance, whether or not the
On Mon, Jun 29, 2009 at 8:17 PM, Nathan Boley wrote:
> On Mon, Jun 29, 2009 at 3:43 PM, Tom Lane wrote:
>> David Fetter writes:
>>> On Mon, Jun 29, 2009 at 01:28:01PM -0700, Nathan Boley wrote:
... They dismiss
singular value decomposition and the discrete wavelet transform as
being
On Mon, Jun 29, 2009 at 3:43 PM, Tom Lane wrote:
> David Fetter writes:
>> On Mon, Jun 29, 2009 at 01:28:01PM -0700, Nathan Boley wrote:
>>> ... They dismiss
>>> singular value decomposition and the discrete wavelet transform as
>>> being too parametric ( which is silly, IMHO )
>
>> Should we have
>> Finally, this creates the partition but ( AFAICT ) it doesn't describe
>> a method for locating the histogram estimate given a point ( although
>> that doesn't seem too difficult ).
> Is that "not difficult," in terms of the math that needs doing, or
> "not difficult," in terms of how well Post
On Mon, Jun 29, 2009 at 06:43:35PM -0400, Tom Lane wrote:
> David Fetter writes:
> > On Mon, Jun 29, 2009 at 01:28:01PM -0700, Nathan Boley wrote:
> >> ... They dismiss singular value decomposition and the discrete
> >> wavelet transform as being too parametric ( which is silly, IMHO
> >> )
>
> >
David Fetter writes:
> On Mon, Jun 29, 2009 at 01:28:01PM -0700, Nathan Boley wrote:
>> ... They dismiss
>> singular value decomposition and the discrete wavelet transform as
>> being too parametric ( which is silly, IMHO )
> Should we have a separate discussion about eigenvalues? Wavelets?
I t
On Mon, Jun 29, 2009 at 01:28:01PM -0700, Nathan Boley wrote:
> > For things like PostGIS, which will want to index in 4 dimensions
> > (x, y, z, t), we might want to have multi-dimensional selectivity
> > histograms and some way to use same.
>
> Another use case is cross column statistics.
Good
> For things like PostGIS, which will want to index in 4 dimensions
> (x, y, z, t), we might want to have multi-dimensional selectivity
> histograms and some way to use same.
>
Another use case is cross column statistics.
> Anybody here qualified to check out this paper on the subject, please
> s
Folks,
For things like PostGIS, which will want to index in 4 dimensions
(x, y, z, t), we might want to have multi-dimensional selectivity
histograms and some way to use same.
Anybody here qualified to check out this paper on the subject, please
speak up :)
http://citeseerx.ist.psu.edu/viewdoc/s
11 matches
Mail list logo
