Errors in Fractional Integer Programming.
Abstract
The generation and propagation of numerical error in the fractional integer program was examined. The developments of error propagation in methods of matrix inversion were partially extended to both the continuous and discrete linear programs. It was shown that, because of the constraints imposed by the linear and integer programs on pivot selection, the strict upper bounds and characterization of error growth in matrix inversion did not apply. It was further shown that the rate of error growth in the integer linear program can be far greater than that for the linear program in continuous variables. The effect of numerical error in decision variables on behavior of an algorithm to solve the fractional integer program was examined. Methods for monitoring and controlling the growth of numeric error in the integer program were examined. The interaction of a group of source-selection rules with a group of dual (entering variable) rules was examined.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1973
- Accession Number
- AD0763206
Entities
People
- John C. Davies