Highs Iterative Solve

[akhilesh-bayer]

I am doing iterative solve in Highs using Python API, where after reading a solution, I am adding constraint with addRows method. However, while solving, it is again loaded into memory where it displays Coefficient ranges:, then solve with existing basis and then simplex iteration starts. However, this process takes some time until simplex iteration starts. Any pointers how I can minimize this time? If I use C++ API, would that be of any help?

Thanks

[jajhall]

If HiGHS is taking time between displaying the coefficient ranges and reporting on simplex iterations, then that's nothing to do with the Python interface. How large is your problem?

I would always recommend using C++ over Python

[akhilesh-bayer]

My problem has 206386 columns, 148294 rows and 570950 non zeros.
Presolve : Reductions: rows 25127(-123167); columns 46657(-159729); elements 191142(-379808)
It takes 37241 simplex iterations to get optimal solution in 36.24 seconds

[jajhall]

Thanks. Presolve has reduced the problem considerably.When you add a row, rather than using presolve and starting from a logical basis HiGHS will solve the unpresolved problem from an advanced basis. Normally this is very much faster. Perhaps you're unlucky.