Re: [Help-glpk] lpx_integer confusion

[Andrew Makhorin]

>Hello, I am attempting to solve a binary
>integer programming problem. I call the lpx_integer function, but
>first I call the lpx_simplex for an initial optimal solution. For
>some reason, occasionally, the lpx_simplex function seems to get some
> information
>in it that makes all further calls to it useless, even when the
> coefficients
>for A and f have changed. The messages I continually get from
> lpx_simplex
>after this point are:
>
>spx_invert: trying to factorize the
>basis using threshold tolerance 0.3
>spx_invert: trying to factorize the
>basis using threshold tolerance 0.7
>spx_invert: the basis matrix is singular
>spx_simplex: initial basis is singular

This means that the initial basis (defined by current statuses of rows
and columns) is singular and therefore cannot be used to start/continue
the search. If you do not intend to provide an initial basis, call
lpx_std_basis or lpx_adv_basis (before lpx_simplex) to construct a valid
initial basis. This might happen if you solved the problem, changed
constraint coefficients at basic variables, and tried to re-optimize it.
On exit lpx_simplex keeps statuses of rows and columns in LPX and uses
them for re-optimization.

>So, if lpx_simplex does not return LPX_E_OK,
>I call lpx_interior instead. lpx_interior seems to do fine, and I
>get the following response:

lpx_integer requires an optimal *basic* solution while lpx_interior
computes an *interior-point* solution.

Andrew Makhorin