Transcript INFORMS05
Forward and Backward
Deviation Measures and Robust
Optimization
Peng Sun (Duke)
with Xin Chen (UIUC) and Melvyn Sim (NUS)
Agenda
Overview of Robust Optimization
Forward and Backward Deviation Measures
Uncertainty Sets and Probability Bounds
Conclusions
Uncertain linear constraints
Uncertain linear constraint
Chance constraint
Hard to solve when is small
Linear constraints with uncertainty
Affine Uncertainty
: zero mean, independent but not necessarily
identically distributed
Overview of Robust Optimization
Worst case
Relies only on the distribution support
Easy to solve (Soyster 1973)
Extremely conservative
Overview of Robust Optimization
Goal of Robust Optimization:
Easy to obtain feasible solutions that satisfy
chance constraint
Not as conservative as worst case
Overview of Robust Optimization
Design of Uncertainty Set
Subset of worst case set, W
: Budget of uncertainty
Control radius of uncertainty set, S
max: Worse case budget
For reasonable probability bound, << max
Overview of Robust Optimization
Robust Counterpart
Semi-infinite constraint
Tractability
Polyhedral Uncertainty Set
Linear Programming (LP)
Conic Quadratic Uncertainty Set
Second Order Cone Programming (SOCP)
Overview of Robust Optimization
Ellipsoidal Uncertainty Set
Ben-Tal and Nemirovski (1997), El-Ghaoui et al.
(1996)
Robust Counterpart is a SOCP
Probability bound for symmetrically bounded
uncertainties:
Overview of Robust Optimization
Polyhedral Uncertainty Set
Bertsimas and Sim (2000)
Robust Counterpart is a LP
Probability bound for symmetrically bounded uncertainties:
More conservative compared to Ellipsoidal Uncertainty Set
Overview of Robust Optimization
Norm Uncertainty Set
Robust Counterpart depends on the choice of norm
Probability bound for symmetrically bounded
uncertainties:
Overview of Robust Optimization
Modeling limitations of Classical
Uncertainty Sets
Uncertainty set is symmetrical, but
distributions are generally asymmetrical
Can we use more information on
distributions to obtain less conservative
solution in achieveing the same probability
bound?
--- Forward and Backward Deviation Measures
Forward and Backward Deviation
Measures
Set of forward deviation measures
Set of backward deviation measures
Alternatively,
Forward deviation
Backward deviation
Forward and Backward Deviation
Measures
Key idea
Important property
Subadditivity
Least conservative forward and backward
deviation measures
Suppose distribution is known
p* = q* if distribution is symmetrical
p*, q* ¸ (standard deviation)
p* = q* = if distribution is Normal
Least conservative forward and backward
deviation measures
p* , q* can be obtained numerically
E.g: p* = q* = 0.58 for uniform distribution in [-1,1]
E.g:
Forward and backward deviation measures
– two point distribution example
Distribution not known
Suppose distribution is bounded in [-1,1]
(but not necessarily symmetrical)
Only know mean and support
Forward and Backward Deviation
Measures: Function g()
General Deviation Measures
General Deviation Measure
Agenda
Overview of Robust Optimization
Forward and Backward Deviation Measures
Uncertainty Sets and Probability Bounds
Conclusions
Uncertainty Sets and Probability Bounds
Model of Data Uncertainty
Recall: Affine Uncertainty
: zero mean, independent but not necessarily
identically distributed
Uncertainty Sets and Probability Bounds
Model of Data Uncertainty
Uncertainty Sets and Probability Bounds
New Uncertainty Set (Asymmetrical)
Uncertainty Sets and Probability Bounds
Recall: Norm Uncertainty Set (Symmetrical)
Uncertainty Sets and Probability Bounds
Generalizes Symmetrical Uncertainty Set
Uncertainty Sets and Probability Bounds
Uncertainty Sets and Probability Bounds
Uncertainty Sets and Probability Bounds
Robust Counterpart
Uncertainty Sets and Probability Bounds
Probability Bound
More information achieve less conservative
solution while preserving the bound. E.g
pj = qj = 1 for symmetric distribution in [-1,1]
pj = qj = 0.58 for uniform distribution in [-1,1]
Conclusions
RO framework
Forward and backward Deviation measures
Affine uncertainty constraints
Independent r.v.’s with asymmetric distributions
Defined from moment generating functions – implying
probability bound
Sub-additivity – linear combinations
Easy to calculate and approximate from support
information
Advantage -- less conservative solution
Next
Stochastic programming with chance constraints