What do these scaling options given by LPsolve.dll mean?

272 Views Asked by At

SCALE_NONE No scaling

SCALE_EXTREME Scale to convergence using largest absolute value

SCALE_RANGE Scale based on the simple numerical range

SCALE_MEAN Numerical range-based scaling

SCALE_GEOMETRIC Geometric scaling

SCALE_CURTISREID Curtis-reid scaling

Additionally, the value can be OR-ed with any combination of one of the following values:

SCALE_QUADRATIC

SCALE_LOGARITHMIC Scale to convergence using logarithmic mean of all values

SCALE_USERWEIGHT User can specify scalars SCALE_POWER2 also do Power scaling

SCALE_EQUILIBRATE Make sure that no scaled number is above 1

SCALE_INTEGERS also scaling integer variables

SCALE_DYNUPDATE dynamic update

SCALE_ROWSONLY scale only rows

SCALE_COLSONLY scale only columns

http://web.mit.edu/lpsolve/doc/set_scaling.htm

By 'scaling', I understand that the program will try to bring all the constraint equations to a a common scale. Say we have, x + y <= 2 and 1000a + 2000b <= 200 then we scale and get x + y <=2 and a + b <= 0.2,