stepwiselm gives different results before and after compilation
    5 views (last 30 days)
  
       Show older comments
    
I have a algorithm / function that i compile into DLL and this DLL is being used by a .Net application as part of a larger project. In this function i call stepwiselm but i am noticing that stepwiselm gives different results when used in matlab vs when compiled as DLL. Moreover this issue does not occur all the time there are some unique sets of input values for which the stepwiselm behaves differently in matlab vs when compiled into DLL. Furthermore this issue occurrs because in the Stepwiselm function i am changing the criteria to  'rquared',  
function [RegCoeff] = DevelopIndividualModel(Input,Output)
Mdl = stepwiselm(Input,Output,'purequadratic','Lower','linear','Criterion','rsquared','PEnter',0.075,'NSteps',100);
RegCoeff = Mdl.Coefficients.Estimate;
end
Try the following values for Input and Output and you will get the same different results between the Matlab function vs Compiled DLL 
Input =   
[  0  250  350  45  50  375 ; 
  75  250  35  45  50  0 ; 
  0  25  35  45  350  0 ; 
  0  25  35  250  50  375 ; 
  0  250  192  250  200  0 ; 
  150  25  192  45  200  375 ; 
  75  25  350  250  350  375 ; 
  150  250  35  250  350  375 ; 
  150  250  350  45  350  0 ; 
  150  175  300  200  50  300 ; 
  125  45  55  75  50  0 ; 
  150  25  350  250  50  0 ]
Output = 
  [  7  
  6  
  6  
  6  
  6  
  6  
  6  
  6  
  9  
  7  
  7 
  7 ]
Output from Matlab is 
RegCoeff = [4.00154247482028, -0.0260252941267720, -0.00474767764661860, 0.00600331079598716, 0.0722078331936164, -0.00201852184261087, -0.0481874258942977, 8.86991940184067e-05, 0.000111244518017688, -0.000251859069586818, 0.000125452183337175, 0]
Output from DLL is 
RegCoeff = [-10.769528534936198,-0.016155202821869604,-0.0084656084656085009,-0.26735445700959604,0.31703485946921428,0.32610476359569857,-0.0011428571428571382,0.00013827160493827187,0.00071347085357696628,-0.0010817811925549459,-0.00082379365502099238,0.0,0.0,0.0]
0 Comments
Answers (0)
See Also
Categories
				Find more on Dimensionality Reduction and Feature Extraction in Help Center and File Exchange
			
	Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!