Prashant-Jonny / accord

Automatically exported from code.google.com/p/accord
0 stars 0 forks source link

Bug in MultivariateLinearRegression Compute() #9

Closed GoogleCodeExporter closed 9 years ago

GoogleCodeExporter commented 9 years ago
What steps will reproduce the problem?
Running a PLS-Regression with 3 dependent variables and 12 predictor variables, 
and then computing the MultivariateLinearRegression obtained from the PLS.

What is the expected output? What do you see instead?
It should compute the regression coheficients correctly, instead it crashes 
with a index out of range exception in the compute method.

What version of the product are you using? On what operating system?
v4.7 on Windows 7

Please provide any additional information below.
It seems you are taking the wrong dimensions in the compute function of the 
MultivariateLinearRegression.
int M = coefficients.GetLength(1); 
it should be:
int M = coefficients.GetLength(0);

Original issue reported on code.google.com by j...@enyetech.com on 29 Jul 2012 at 6:01

GoogleCodeExporter commented 9 years ago
Sorry, my bad! I had called the PLS constructor with 
PartialLeastSquaresAnalysis(DependentVars, PredictorVars, ...) instead of 
(PredictorVars, DependentVars, ...)

Original comment by j...@enyetech.com on 29 Jul 2012 at 6:23

GoogleCodeExporter commented 9 years ago
Thanks for the bug report. No worries. So, is it working now? 

I will be improving the argument validation in those methods.

Original comment by cesarso...@gmail.com on 29 Jul 2012 at 5:23

GoogleCodeExporter commented 9 years ago
It's working like a charm now.
Im trying out most of the statistical and machine learning classes, and so far 
so good.

Thanks for such a great framework!
(By the way, do you have some kind of roadmap for Accord?)

Original comment by j...@enyetech.com on 29 Jul 2012 at 6:37

GoogleCodeExporter commented 9 years ago
The newest release provides more argument checks in PLS. 

The current goal of the framework is to support my own field research (in 
computer vision and machine learning). However, this does not means other 
features which I won't be directly using will be left aside. If you have 
suggestions for a new implementation or would like to contribute, feel free to 
submit an enhancement request or a code contribution.

Best regards,
Cesar

Original comment by cesarso...@gmail.com on 1 Sep 2012 at 9:44