In Practical Linear Programming we looked at a very small LP model. We could build that model directly using a text editor because it was so small. It represented one element of a model of an oil refinery. What if we had been asked to build a model of the entire oil refinery? How should we have approached the task? There are really two separate issues here: - how should we set about thinking about our model?
- how should we implement it, i.e. what software should we use?
We have seen in the last lecture several different ways of implementing the same problem using different types of software. We shall see shortly be using another one, the algebraic modelling language XPRESS. As we have discussed, there is no single "right" piece of software which is appropriate everywhere. Much will depend on the software environment in which the MP model is to lie. The second question is therefore essentially pragmatic. It can be answered by those engaged in implementing the model. What really matters is that the software should not dictate how the problem is to be solved. We should decide what our problem is and how we will tackle it. We should consider the general capability of the software which is available. But we should not let the particular details of software packages fetter the analysis of our problem. It is the first question, therefore, which is the important one. From the 1950s to the 1970s Mathematical Programming was such an intensive user of computing power that the reverse was true. People really did think in the terms which the software packages dictated. This lecture sets out recommendations for how to think about MP problems. In the tradition of mathematicians, the only prerequisites are pencil and paper. |

LP Training > 3. Algebraic Formulation >