fboehlandt
Active Member
- Joined
- Sep 9, 2008
- Messages
- 334
Hi everyone,
I'm currently working on an optimization problem with equaliy and inequality constraints. The objective is to minimize the sum of squares in a multifactor regression model where the coefficients are subject to the following constraints:
- sum of coefficients = 1 (equality constraint)
- non-negativity (inequality constraint)
I calculated the coefficients using both Solver as well as the Karush-Kuhn-Tucker algorithm to find the solution to this non-linear programming problem. The latter method requires some extensive calculations with matrices. Unfortunately, the resulting coefficients differ with respect to the third position after the decimal point. I was wondering if this might be attributable to some lack of precision when multiplying large matrices in Excel, a different approach by Solver optimizing non-linear functions, or whether I'm making some crucial mistake calulating the coefficients manually. Does anybody have any experience solving non-linear programming problems manually?
p.s. I will gladly provide my model calculations if required
I'm currently working on an optimization problem with equaliy and inequality constraints. The objective is to minimize the sum of squares in a multifactor regression model where the coefficients are subject to the following constraints:
- sum of coefficients = 1 (equality constraint)
- non-negativity (inequality constraint)
I calculated the coefficients using both Solver as well as the Karush-Kuhn-Tucker algorithm to find the solution to this non-linear programming problem. The latter method requires some extensive calculations with matrices. Unfortunately, the resulting coefficients differ with respect to the third position after the decimal point. I was wondering if this might be attributable to some lack of precision when multiplying large matrices in Excel, a different approach by Solver optimizing non-linear functions, or whether I'm making some crucial mistake calulating the coefficients manually. Does anybody have any experience solving non-linear programming problems manually?
p.s. I will gladly provide my model calculations if required