1 / 35

Non-Linear Programming

Non-Linear Programming. Same structure variables objective function constraints No restrictions Except typically variables must be continuous. Examples. How to model binary variables x is 0 or 1 Equivalent continuous formulation x(1-x) = 0 NOT LINEAR!. Location Problem. Variables

cassia
Télécharger la présentation

Non-Linear Programming

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Non-Linear Programming • Same structure • variables • objective function • constraints • No restrictions • Except typically variables must be continuous

  2. Examples • How to model binary variables • x is 0 or 1 • Equivalent continuous formulation • x(1-x) = 0 NOT LINEAR!

  3. Location Problem Variables x is the X coordinate of the facility y is the Y coordinate of the facility Objective Minimize Distance traveled to deliver goods Constraints - None

  4. Formulation • minimize 200*sqrt((x-5)2 + (y-10)2) + 150*sqrt((x-10)2 + (y-5)2) + 200*sqrt((x-0)2 + (y-12)2) + 300*sqrt((x-12)2 + (y-0)2)

  5. Pooling Problem • Blend crudes in pools • Blend Alaska 1 and Alaska 2 • Make products from the pools • Regular • Unleaded • Premium • Composition constraints on final products • Premium  2.8% Sulfur  90 Octane • Sells for $0.86/gal, minimum 5000 gals

  6. Diagram Lead Premium Alaska 2 Unlead Alaska Pool Texas Reg. Alaska 1

  7. Input Variables • Lead - gallons daily • LeadPrem - gallons of lead used in premium daily • LeadReg - gallons of lead used in regular daily • Alaska - gallons of Alaska pool daily • Alaska1 - gallons of Alaska 1 used in pool daily • Alaska2 - gallons of Alaska 2 used in pool daily • AlaskaPrem - gals of Alaska pool used in prem. daily • AlaskaReg - gals of Alaska pool used in reg. daily • AlaskaNoL - gals of Alaska pool used in No lead daily • Texas - gallons of Texas used daily • TexasPrem - gals of Texas used in prem. daily • TexasReg - gals of Texas used in reg. daily • TexasNoL - gals of Texas used in No lead daily

  8. Output Variables • Prem - Gals of Premium produced daily • Reg- Gals of Regular produced daily • NoL - Gals of No Lead produced daily

  9. Composition Variables • For convenience • AlaskaSulfur - sulfur content of Alaska pool • AlaskaOctane- octane of Alaska pool

  10. Constraints • Define Alaska Pool • Alaska = Alaska 1 + Alaska 2 • AlaskaSulfur = (4%*Alaska 1 + 1% * Alaska 2)/Alaska • AlaskaOctane=(91*Alaska 1 + 97*Alaska 2)/Alaska • Use Alaska Pool • Alaska = AlaskaPrem + AlaskaReg + AlaskaNoL

  11. Constraints Cont’d • Define Products • Prem = AlaskaPrem+ TexasPrem + LeadPrem • Reg = AlaskaReg+ TexasReg + LeadReg • NoL = AlaskaNoL+ TexasNoL • Constrain Composition • AlaskaSulfur*AlaskaPrem + .02*TexasPrem  .028*Prem • AlaskaSulfur*AlaskaNoL + .02*TexasNoL  .03*NoL • AlaskaSulfur*AlaskaReg + .02*TexasReg  .03*Reg • AlaskaOctane*AlaskaPrem + 83*TexasPrem  94*Prem • AlaskaOctane*AlaskaNoL + 83*TexasNoL  88*NoL • AlaskaOctane*AlaskaReg + 83*TexasReg  90*Reg

  12. Constrain Volumes • Prem  5000 • Reg  5000 • NoL  5000 • Upper Limits • Texas  11000 • Lead  6000

  13. Objective • Maximize Profit • Revenues from Products • 0.86*Reg + 0.93*NoL + 1.06*Prem • Costs of Raw Materials • 0.78*Alaska 1 + 0.88*Alaska 2 + 0.75*Texas + 1.30*Lead

  14. Formulating NLPs • As in the book • No need for abstraction • Some off the shelf software (MINOS) • Requires more sophistication to use • Does not typically provide guarantees

  15. Getting Guarantees • When we can use an LP formulation with a non-linear objective • Minimize Cost and things get more expensive as we get more • Maximize Profit and profits decrease as we sell more

  16. Minimize Cost Total Cost Volume Purchased

  17. Minimize Cost Total Cost Volume Purchased

  18. Easy Problem • The Cost Function lies below the linear approximation • No incentive to use any weights other than consecutive ones • Don’t need Integer Programming

  19. Convex Function • Lies below the line • Technically: A convex function has the property that for each pair of points x and y and weight w between 0 and 1 the function evaluated at wx + (1-w)y (a fraction w of the way from y towards x) is wf(x) + (1-w)f(y) (the same fraction of the way from f(y) towards f(x))

  20. Minimize Cost Total Cost Volume Purchased

  21. Convex in 2 dimensions

  22. Maximize Profit Total Profit Volume Sold

  23. Maximize Profit Total Profit Volume Sold

  24. Easy Problem • The Profit Function lies above the linear approximation • No incentive to use any weights other than consecutive ones • Don’t need Integer Programming

  25. Concave Function • Lies above the line • Technically: A concave function has the property that for each pair of points x and y and weight w between 0 and 1 the function evaluated at wx + (1-w)y (a fraction w of the way from y towards x) is  wf(x) + (1-w)f(y) (the same fraction of the way from f(y) towards f(x))

  26. Maximize Profit Total Profit Volume Sold

  27. What makes these easy • With no constraints • Local Optimum is best in a small neighborhood, e.g., as good as every point within epsilon of it. • Convex minimization: A local optimum is a global optimum, e.g., a best answer • Concave maximization: A local optimum is a global optimum.

  28. Tough Problems Local Max Local Max Local Min Local Min

  29. Convex Sets • A set with the property that for every pair of points in the set, the line joining the points is in the set as well is a CONVEX SET • Points in a convex set can see each other

  30. Convex Sets • Linear Programming Feasible regions

  31. Non-convex Sets • Feasible Region of Integer Programs

  32. Easy Problems • Convex Minimization over a convex set • Objective is a convex function • Constraints define a feasible region that is a convex set • Any Local minimum is a global minimum

  33. Easy Problems • Concave Maximization over a convex set • Objective is a concave function • Constraints define a feasible region that is a convex set • Any Local maximum is a global maximum

  34. Non-convex Sets are Hard Feasible Profit Volume Sold

More Related