12:00, 24 Apr 1996, WeH 7220


      Linear Programming Methods for Stochastic Dynamic Programs
			  Mike Trick (GSIA)

Linear programming is a well-known and often rejected technique for
solving large-scale stochastic dynamic programs.  Recent advances in
linear programming implementations make this technique competitive
with standard techniques such as value iteration.  Furthermore, a
close examination of dual values makes it possible to adaptively
refine discretizations of continuous models.  Finally, this technique
can be extended to spline approximations with similar computational
advantage.  This work has been done jointly with Stan Zin, also of
GSIA.