Prev: Read data set metadata from HDF4 files?
Next: parfor with sedumi or yalmip
From: rams on 31 Mar 2010 10:37
Hi Greg,

Thanks for the reply....as i am a beginner to this kind of problem, could you explain the more detailed procedure to attack this problem....
From: Greg Heath on 31 Mar 2010 21:59
On Mar 31, 2:37 pm, rams <lrams...(a)gmail.com> wrote:
> Hi Greg,
>
> Thanks for the reply....as i am a beginner to this kind of problem, could you explain the more detailed procedure to attack this problem....

1. You want to use N data points to create a model

y = f(x,P),

x matrix of input vectors, size(x) = [N 5]
P parameter matrix, size(P) = model dependent
y vector of scalar outputs, size(y) = [N 1]

I would try linear, quadratic and neural network models.

2. Using f and P you want to create an inverse model

x = g(y,P).

I don't know how to do part 2 without more information.

3. a. Look at the 5 plots x(i) vs y. Are any of them monotonic?
b. Are any of the 5 x(i) redundant (i.e., can be estimated
from the other 4)?
c. Are any of the 5 x(i) irrelevant (i.e., can be eliminated
in estimating y)?

Hope this clarifies my previous reply.

Greg

From: rams on 1 Apr 2010 12:10
I have 664 data points so that in this case N=664
i have 5 independent variables with 664 values in each and they form a matrix of size [664 5]. Could you please tell me how do i choose parameter matrix p for both linear and quadratic models( and models too).
From: Greg Heath on 1 Apr 2010 22:06
On Apr 1, 4:10 pm, rams <lrams...(a)gmail.com> wrote:
> I have 664 data points so that in this case N=664
> i have 5 independent variables with 664 values in each and they form a matrix of size [664 5]. Could you please tell me how do i choose parameter matrix p for both linear and quadratic models( and models too).

For models that are linear in variables and coefficients
you have your choice of
1. Backslash
doc slash
help slash
2. Regress
doc regress
help regress
3. Stepwisefit
doc stepwisefit
help stepwisefit

I suggest you use all 3 and compare before
continuing.

For models that are nonlinear in variables but linear
in coefficients (e.g., polynomials) you can still
use the above three approaches.

Search on

greg heath vectorization quadratic

to see how to set up the quadratic model that is
linear in coefficients.

Hope this helps.

Greg
First  |  Prev  | 
Pages: 1 2 3
Prev: Read data set metadata from HDF4 files?
Next: parfor with sedumi or yalmip