multivariate production systems optimization - Stanford University
multivariate production systems optimization - Stanford University
multivariate production systems optimization - Stanford University
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
The second step of the Greenstadt (1967) modification, replacing small eigenvalues<br />
with infinity, is based on the objective function being insensitive to any small eigenvalues.<br />
Since the objective function is insensitive to a parameter with a small eigenvalue, this<br />
parameter should not be allowed to influence the step-direction. This is precisely what is<br />
achieved by replacing the eigenvalue with infinity. Replacing an eigenvalue with infinity is<br />
accomplished by replacing the reciprocal of the eigenvalue with zero when inverting the<br />
Hessian.<br />
Conceptually, the Greenstadt (1967) modification applied to Newton’s Method<br />
should result in a robust algorithm that is rapidly convergent and not affected by insensitive<br />
parameters; the caveat being that it is very expensive computationally. For information on<br />
the performance of the Newton-Greenstadt algorithm applied to petroleum engineering<br />
problems, see Barua, Horne, Greenstadt, and Lopez (1989).<br />
6.3 Quasi-Newton Methods<br />
Newton’s Method is quadratically convergent when starting with a good initial estimate.<br />
However, it is very expensive since the Hessian must be built and solved every iteration,<br />
particularly when the Hessian is built with finite difference approximations. The idea<br />
behind Quasi-Newton methods is to compromise on the speed of convergence while saving<br />
on the expense associated with building and solving the Hessian. Instead of building and<br />
solving an exact Hessian every iteration, Quasi-Newton methods attempt to update an<br />
approximation of the Hessian.<br />
Quasi-Newton theory is based on multidimensional generalizations of the secant<br />
method. The object is to build up secant information as the iterations proceed. Suppose at<br />
the kth iteration, the Newton step {xk+1 - xk } causes a change in the gradient of {gk+1 -<br />
gk }. Then the next Hessian approximation will satisfy the secant passing through these<br />
two iterates if<br />
xk+1 - xk = Hk+1 gk+1 - gk<br />
71<br />
(6.28)<br />
This condition is termed the Quasi-Newton condition and is the design criteria for Quasi-<br />
Newton methods. Notice the similarity between this condition and the Newton condition,<br />
Equation 6.12. This condition forces the Hessian approximation to exactly match the<br />
gradient of the function in the displacement direction, {xk+1 - xk }. For multiple