Uploaded a new snapshot (rev. 100) for public consumption. In this release I’ve modified the model building algorithm – the algorithm responsible for converting a model in the spreadsheet form into a form the backend optimization algorithm can digest – to make it more robust, but this improvement is still not enough to solve those test files sent by Ludovic and Markus. Assuming that their models are actually feasible (I know at least Markus’ model is feasible as he also included Excel Solver’s solution), I have at least two more things in my mind that need to be done before my solver can solve their models.

Alternatively, I could use another variant of simplex that’s especially suited for the type of models they sent me – the type that has a specific upper or lower bound for each decision variable. But the code I wrote for the upper-and-lower-bounded simplex still has a few bugs to be sorted out before it can be plugged in for use. Not sure what the best course of action is…

*Related*

## Published by Kohei Yoshida

LibreOffice contributor, spreadsheet nerd, and machine learning beginner.
View all posts by Kohei Yoshida