Pyomo / mpi-sppy

MPI-based Stochastic Programming in PYthon
https://mpi-sppy.readthedocs.io
Other
65 stars 38 forks source link

Updates to proximal approximation #368

Closed bknueven closed 6 months ago

bknueven commented 6 months ago

The current proximal approximation approach does not allow you to start with no cuts.

This PR would update the proximal approximation code such that:

  1. Allows for starting with an empty set of cuts
  2. Makes starting with the empty set of cuts the new default
  3. Adds two cuts per pass: the "nearest point" cut (i.e., at x = argmin [( x - \hat{x} )^2 + ( x^2 - \hat{y} )^2] and the "straightforward projection cut" (i.e., at x = \hat{x}). Before we just added the "nearest point" cut.
bknueven commented 6 months ago

After some more extensive testing, it isn't clear this is any better than the current implementation. Closing.