Originally posted by: GiovanniP

Hello,

Consider an LP. Given an optimal basis B, I was wondering:

1) if there is a way to access B^-1 (which I really doubt) and

2) what do you think is the closest way to perform B^-1b, given a new right hand side b, and check if the basis is still feasible (i.e. if the product is positive)? Is it by accessing sensitivity analysis information?

Thanks for any advice

I think you are looking for functions CPXinvcol() and CPXinvrow()?

Originally posted by: T_O

Concering your first question: Look at CPXXbinvcol and CPXXbinvrow. (Daniel forgot a "b", links were correct.)

Concering your second question, I'm not sure what the best way is. Maybe change b and do a warmstart with dual simplex and iteration limit 1?

Best regards,
Thomas

Originally posted by: GiovanniP

Thank you for your quick answer. It is actually what I was looking for.

I am having a hard time finding the equivalent in terms of Java API. Any clue?

Best regards,

Giovanni

These things don't exist in the Java API.

The Java API implements Concert Technology and Concert Technology does not have a matrix centered view on the problem. Hence those methods are not available in the Java API.