SQG

Uploaded from authorPOINTLite
Views:
 
Category: Education
     
 

Presentation Description

No description available.

Comments

Presentation Transcript

Stochastic Quasi-Gradient Methods: 

Stochastic Quasi-Gradient Methods Roger J-B Wets University of California, Davis February 15, 2005

Stochastic optimization: 

Stochastic optimization Formulation Properties: S

Subgradients of convex fcns: 

Subgradients of convex fcns

Minimization algorithms: 

Minimization algorithms Step type 1

Minimization algorithms: 

Minimization algorithms Step type 2 proj

“repeated” projections: 

“repeated” projections Convex program: quadratic objective function quadratic program if S is a polyhedral set Many applications: projection is a simple/efficient non-negative, convex, bounded away from 0

SQG Iterates: 

SQG Iterates basic strategy:

SQG: Stochastic Optimization: 

SQG: Stochastic Optimization . sqg: justification:

SQG: Stochastic Optimization: 

SQG: Stochastic Optimization . value estimate: justification:

A (simple) location problem: 

A (simple) location problem Pop. Size of 12 districts: 11 # 26. Probabilistic choice of shopping district: shortage cost: 4, holding cost: 0.5 (excess) decision: location of facilities (shopping malls)

“preferences” table: 

“preferences” table

Formulation: 

Formulation from objective: probability of sample determined by customer behavior

Objective Value: iterates : 

Objective Value: iterates Estimate of the objective per iterate

Objective Value (2): iterates : 

Objective Value (2): iterates Estimate of the objective per iterate Facilities: 18.57 15.90 19.13 16.35 27.25 20.75 21.88 17.81 19.11 17.52 18.62 19.60 Distr.Pop: 14 11 14 13 26 23 22 11 14 12 18 10

Objective Value (3): iterates : 

Objective Value (3): iterates Facilities: 24 22 23 20 26 22 23 22 22 20 22 25 : 271 Distr.Pop: 19 16 19 16 27 21 22 18 19 18 19 20 : 234

a.s. Convergence: 

a.s. Convergence For now presumed optimal sol’n at iteration  projection implies:

a.s Convergence: 

a.s Convergence taking condition expectation w.r.t. F assumption(a.): with

a.s Convergence: 

a.s Convergence Hence Assumption(b.): where with

a.s. Convergence: 

a.s. Convergence recursively from (a)

a.s. Convergence: 

a.s. Convergence Thus assumption (c.) and there exists a subsequence such that

Review of assumptions: 

Review of assumptions (a.) (b.) (c.)

“stumbling” blocks: 

“stumbling” blocks Projection Step size: adaptive, adjust (increase, decrease) based on the variance of the stochastic quasi-gradient Stopping criterion: like for step-size, but more generally comparison of the values of the objective:

A short history: 

A short history Stochastic approximation methods Robbins & Monro, Kiefer & Wolfowitz (‘50) SQG: Theory Shor, Poljak, Ermoliev, Fabian (‘60), Kushner(‘70),Pflug, Ruszczynski (‘80), Implementation: Gaivoronski, Gupal, Norkin (‘80 … 2005)