This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: IEOR E4007 G. Iyengar October 8th, 2008. Homework #3 Due : October 22nd, 2008. 1. Exercise 3.15 from Optimization methods for finance In this problem there is a discrepancy between the answers generated by EXCEL and AMPL/CPLEX. The issue is that there are many (at least two!) optimal corner solutions, i.e. two optimal bases. AMPL/CPLEX finds one and EXCEL finds the other. Since the sensitivity report depends on the dual variables which themselves are a function of the optimal basis, we end up with two different sensitivity reports. If you solve this problem using AMPL/CPLEX we get the following sensitivity report: x x.rc x.down c x.up 4 1 1.5 1 7 0.666667 1 1 2 1 1 1 1.5 3 7 1 1 4 1 1 1e+20 5 3 1 1.5 6 0.333333 0.666667 1 1e+20 req.dual req.down req.current req.up 0.333333 12.5 14 15.5 11e+20 13 14 2 0.333333 13.5 15 25.5 31e+20 16 19 4 0.333333 16 19 22 5 0.333333 16.5 18 19.5 6 10 11 12 The AMPL code shift.mod and shift.run that generates this report is posted on the web. Note that the days are numbered from 0,1,...,6 with Monday labeled , Tuesday labeled as 1 , etc.. (a) The total number of workers = 4 + 7 + 1 + 4 + 3 + 3 = 22. This is same regardless of how you solve the problem. (b) The Thursday constraint is slack and the allowable increase is 3 . Therefore, there is no change in the number of workers needed. 1 (c) The dual variable corresponding to the Monday constraint is 1 3 and the change in the requirement is 3 which is well within the allowable decrease for the EXCEL (i.e. Text) sensitivity report. Therefore, the net change in the objective is 3 1 3 = 1. For the AMPL/CPLEX report the allowable decrease is not within the RHS range. Therefore, we can only say that the decrease in the objective is within [ 1 , . 5]. (d) If the solver is EXCEL (as is the case with the book!) then the sensitivity report in the text is correct. In this case the increase of 0 . 1 is within the allowable range for the objective coefficient of the Shift4...
View
Full
Document
 Summer '09
 OptimizationModelsandMethods
 Optimization

Click to edit the document details