Sum of weight times each sparse vector in the declared coordinates.
Computed complete pricing gap passed the target and simplex sum passed its explicit Float64 summation allowance.
Computed nonnegative pricing-slack sum plus gapMassAllowance; not an interval bound on dot-product roundoff.
Absolute measured (sum(weights)-1)*minimumPrice correction in objective units.
Actual pairwise weight transfers performed.
Minimum a_i dot aggregate + offset_i over every atom.
Half squared aggregate norm plus the weighted affine coefficient.
Absolute measured sum(weights)-1 after normalization and complete pricing.
Gap termination, explicit work/iteration exhaustion or unrepresentable progress.
Nonnegative weights summing to one to floating arithmetic accuracy.
Conservative scalar visits charged before the corresponding operation.
Owned convex simplex iterate with a complete Frank-Wolfe gap certificate.