ISYE6669 LP 10 20 1 - AndySun - Corrected

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Deterministic

Optimization
Large-Scale Optimization

Andy Sun
Assistant Professor
Stewart School of Industrial and Systems Engineering

Example for Column Generation


Large-Scale Optimization

Learning Objectives for this module

• Apply column generation algorithm


to numerically solve a cutting stock
problem
Column Generation Algorithm
We want to solve a large-scale LP with many variables
1. Pick a subset 𝐼 of columns {𝐴𝑗 } and variables 𝑥𝑗 for 𝑗 ∈ 𝐼 ⊆ {1,2, … , 𝑁}
2. Solve (RMP) with a subset of columns 𝐼

Solving (RMP) is not challenging (and it shouldn’t be) because of the small
number of columns.
Column Generation Algorithm (cont.)
3. Pricing probem

– If 1 − 𝑍መ ≥ 0, all reduced costs are nonnegative. Terminate.


– If 1 − 𝑍መ < 0, add the generated column [𝑎1 ; 𝑎2 ; 𝑎3 ] to (RMP). Go to
Step 1.
Numerical Example
• Large roll width 𝑊 = 100
• Demand: 𝑤1 = 25, 𝑏1 = 150; 𝑤2 = 35, 𝑏2 = 200; 𝑤3 = 45, 𝑏3 = 300
• Patterns
Iteration 1: Solve RMP
• Initial set of patterns and RMP

• Three demands (linear equality constraints) ➔three patterns (columns)


• Optimal solution: 𝑥ҧ9 = 150, 𝑥ҧ10 = 200, 𝑥ҧ11 = 300
• Optimal basis:
Iteration 1: Pricing Problem
• Form and solve knapsack problem
Iteration 2: RMP
• Set of columns 𝐼 = 1,9,10,11
• Solve RMP

• Optimal solution: 𝑥ҧ1 = 37.5, 𝑥ҧ9 = 0, 𝑥ҧ10 = 200, 𝑥ҧ11 = 300


• Optimal basis:
Iteration 2: Pricing Problem
• Compute minimum reduced cost
Iteration 3: RMP
• Set of columns 𝐼 = 1,6,9,10,11
• Solve RMP

• Optimal solution: 𝑥ҧ1 = 12.5, 𝑥ҧ6 = 100, 𝑥ҧ9 = 0, 𝑥ҧ10 = 0, 𝑥ҧ11 = 300
• Optimal basis:
Iteration 3: Pricing Problem
• Compute minimum reduced cost
Iteration 4: RMP
• Set of columns 𝐼 = 1,6,9,10,11,14
• Solve RMP

• Optimal solution: 𝑥ҧ1 = 12.5, 𝑥ҧ6 = 100, 𝑥ҧ9 = 𝑥ҧ10 = 𝑥ҧ11 = 0, 𝑥ҧ14 = 150
• Optimal basis:
Iteration 4: Pricing Problem
• Compute minimum reduced cost
Summary

• We learned:
– The detailed steps of applying column
generation to the cutting stock problem

You might also like