Prev: Matlab version for Windows Vista and Windows 7
Next: Types of video file formats in Mat lab
From: Rupika Bandara on 12 Jul 2010 09:11
Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <i1f36i$sdg$1(a)fred.mathworks.com>...
> On 7/10/2010 6:45 PM, Rupika Bandara wrote:
> > Should I multiply 150x150 matrix by matrix a (16*150) to get values
> > according to 16 observations? Otherwise, how can I find values
> > corresponding to 16 observations separately?
>
> Rupika, it may be that you want the _second_ output from PRINCOMP, but
> it's pretty hard to tell from your description.

I use 16x 150 matrix to find PCA values. 16 are no of observations (no of damage cases of the beam) and 150 are dimension of the observations. Then using princomp function I can find most prominant PCA values. It is a 150x150 matrix. Then I want to use PCA values to neural network as input values. But I need to find PCA values separately according to 16 damage cases from 150x150 matrix. How can I find values ?
Your help is highly appreciated.
From: Peter Perkins on 12 Jul 2010 12:51
On 7/12/2010 9:11 AM, Rupika Bandara wrote:
> I use 16x 150 matrix to find PCA values. 16 are no of observations (no
> of damage cases of the beam) and 150 are dimension of the observations.
> Then using princomp function I can find most prominant PCA values. It is
> a 150x150 matrix.

It isn't. That's the matrix of _all_ the principal component
coefficients. You need to use the third output of PRINCOMP to select
the number of PCs that you want. I recommend you read over this section
of the User Guide:

<http://www.mathworks.com/access/helpdesk/help/toolbox/stats/brkgqnt.html#f75476>
From: Rupika Bandara on 12 Jul 2010 20:06
Peter Perkins <Peter.Perkins(a)MathRemoveThisWorks.com> wrote in message <i1fh7e$dn$1(a)fred.mathworks.com>...
> On 7/12/2010 9:11 AM, Rupika Bandara wrote:
> > I use 16x 150 matrix to find PCA values. 16 are no of observations (no
> > of damage cases of the beam) and 150 are dimension of the observations.
> > Then using princomp function I can find most prominant PCA values. It is
> > a 150x150 matrix.
>
> It isn't. That's the matrix of _all_ the principal component
> coefficients. You need to use the third output of PRINCOMP to select
> the number of PCs that you want. I recommend you read over this section
> of the User Guide:
>
> <http://www.mathworks.com/access/helpdesk/help/toolbox/stats/brkgqnt.html#f75476>


Thank you for your help. Third output is variance and using it I can find most prominent PCA values of the data set. Then I want to know whether the PCA of first row of 150x150 matrix is corresponding to first damage case or not. Likewise, first 16 rows out of 150 rows are corresponding to 16 damage cases or not. Please help me.
First  |  Prev  | 
Pages: 1 2
Prev: Matlab version for Windows Vista and Windows 7
Next: Types of video file formats in Mat lab