Skip to content

Latest commit

 

History

History
37 lines (25 loc) · 2.33 KB

README.md

File metadata and controls

37 lines (25 loc) · 2.33 KB
Raw

lroaz

Implementation of a Phi-Divergence Two-Stage Linear Program with Recourse (PhiLP-2) for Tucson water data. This project originally started life as an implementation fo the Likelihood Robust Two-Stage Linear Program with Recourse, but was later generalized to include other phi-divergences as well.

Classes are as follows:

  • @LPMODEL contains the structure of a two-stage stochastic linear program with recourse.
  • @LRLP implements the modified Bender's Decomposition used to solve the PhiLP-2 model, formerly the Likelihood Robust model.
  • @PhiDivergence is a class to contain the phi-divergences and their important properties.
  • @Solution class is used to store the candidate and best solution as the modified Bender's Decomposition algorithm runs.

A few functions of interest:

  • SolveLRLP.m implements the basic Bender's Decomposition loop using the LRLP class
  • GatherLROData.m loops over a variety of parameter values to collect data on optimal solution of the PhiLP-2 problem.

=====

DR. LANSEY'S STUDENTS:

For the code to build an LP based on the excel spreadsheets, you want the LPModel class, contained in the folder @LPModel

The file OneStageAZ.m shows you the basic use of the LPModel class.

The doc folder contains some technical documentation on how LPModel constructs the water model.

I currently have LPModel set to never write to Excel (it doesn't work on Linux). To enable writing to excel, open @LPModel/LPModel.m, and set the variable writeToExcel = true. This is line 52 as of the time I'm writing this.

For reference, the files in LPModel:

  • LPModel.m contains the basic information about LPModel--data and some small methods.
  • BuildA.m builds the submatrices A, A_st and A_log. Alicia can tell you all about those.
  • ExpandA.m creates the full constraint matrix out of A, A_st and A_log.
  • BuildVectors.m Builds the cost, right hand side, and lower and upper bound vectors. It should work for both one and two stage problems. NOTE BuildVectors will only work for Time_Lag = 1 or Time_Lag = 12. This is based on what I see in sf.xlsx
  • GenerateDiagonal.m is used by ExpandA.m to build the full constraint matrix. Don't modify this file.
  • PrintConstraint.m prints the constraints in a human readable format, which should be helpful for debugging any changes to the code, or creating new spreadsheets.