Sparse algebra interface and lambda .= 0 at first iteration about hiop HOT 4 CLOSED

Works like it's supposed to, but will change the starting procedure to call the Hessian a bit later in the initialization process, which will make lambdas nonzero in most of the cases.

from hiop.

I pushed the fix see first commit in PR#96. However this will not fix your problem, as Newton IPM does not have the LSQ initialization for the multipliers. Without this they will still be zero when Hessian evaluation is reached.

Easy fix for now: hiopInterfaceMDS::starting_point support dual restart so use it to initialize all three multipliers arrays to 1.

from hiop.

let me know if starting_point solution works and will close the issue as the LSQ initialization will be handled in issue #97

from hiop.

closing this as issue #97 covers the require developments to fix this issue.

from hiop.