Prev: Initial Output of Merge Bloc when fed by buses
Next: Database Exec
From: kk KKsingh on 7 Jul 2010 12:48
Some time When i test my algorithm which involve inversion using least squares ! I get error my matrix is badly condition.....Whats best remedy for this

Thanks

Kumar
From: John D'Errico on 7 Jul 2010 12:58
"kk KKsingh" <akikumar1983(a)gmail.com> wrote in message <i12b44$831$1(a)fred.mathworks.com>...
> Some time When i test my algorithm which involve inversion using least squares ! I get error my matrix is badly condition.....Whats best remedy for this
>

Best remedy?

1. Don't use a matrix inverse in the first place.

help slash
help pinv

2. Learn how to solve least squares problems. (Read a good
book on the subject. Or take a class.)

3. Learn about scaling your data.

4. Choose better models, that don't have numerical problems
(Polynomial models are notorious for conditioning problems,
because people abuse the process, fitting too high an order
model.)

Learn to use the tools that you will use.

John
From: Greg Heath on 7 Jul 2010 20:58
On Jul 7, 12:48 pm, "kk KKsingh" <akikumar1...(a)gmail.com> wrote:
> Some time When i test my algorithm which involve inversion using least squares ! I get error my matrix is badly condition.....Whats best remedy for this

The bottom line basics:

Avoid predictors that are poorly scaled and/or
highly correlated wirh other predictors.


Hope this helps.

Greg
 | 
Pages: 1
Prev: Initial Output of Merge Bloc when fed by buses
Next: Database Exec