Title: Robust Optimization Concepts and Examples
1Robust OptimizationConcepts and Examples
- Yuriy Zinchenko
- Shane G. Henderson
- ORIE, Cornell University
2Outline
- What can go wrong with LP?
- A familiar blend problem
- The general picture
- Robust linear programming
- Software, resources, practicalities
- Radiation therapy for cancer treatment
3What can go wrong with LP?
- Tough LP problem
- max x y
- s/t 1 x ? 1 1 y ? 1
- x, y ? 0
4Blend Problem
- blend to get output properties at minimum cost
for anyinput propertieswithin reason
5Blend constraints
- Typical constraint looks like
- Low 10 x1 12 x2 7 x3 High
- Changes to
- Low a1 x1 a2 x2 a3 x3 High
- for any vector a that is close to (10, 12, 7)
6General robust LP
- min cTx
- s/t A(1) x ? b1
- A(2) x ? b2
- A(3) x ? b3
7A more detailed view
- Simple linear constraint
- a x ? 1
- x ? 0
- with a close to 1, namely
- 0 ? a ? 2
- Want x to work for all such a
- How do we deal with it?
8- a x ? 1, x ? 0 for all 0 ? a ? 2
- max a x ? 1, x ? 0
- 0 ? a ? 2, x ? 0
- 2 x ? 1 , x ? 0
- x ? 1/2 , x ? 0
9- A slightly more involved example
- a x b y ? 1
- where (a, b) close to (1, 1), namely in
- Ellipsoidal (spherical)
- uncertainty set U
- (a, b) is in U if
- (a, b) (a0, b0) (Da, Db)
- with (a0, b0) (1, 1) and Da2 Db2 ? 1
10- Ellipsoidal uncertainty set U
- (a, b) (a0, b0) (Da, Db)
- (a0, b0) (1, 1)
- Da2 Db2 ? 1
- Want (x, y) to satisfy
- a x b y ? 1,
- for all (a, b) from U
11- a x b y ? 1 for all (a, b) in U
- max a x b y ? 1
- (a, b) in U
- What can we say about a x b y ?
- a x b y (a0 Da) x (b0 Db) y
- (a0 x b0 y) (Da x Db y)
12- For a moment, think of (x, y)
- as your objective function (fixed)
- max a x b y (? 1 ?)
- (a, b) in U
- same as
- (a0 x b0 y) max (Da x Db y) (? 1 ?)
- Da2 Db2 ? 1
13- max (Da x Db y) (? 1 - (a0 x b0 y) ?)
- Da2 Db2 ? 1
- Here
- Da x Db y
- ? (x, y)
- (x2 y2)1/2
- the length
- of (x, y)
U
(a0, b0)
14- a x b y ? 1 for all (a, b) in U
- max a x b y ? 1
- (a, b) in U
- (a0 x b0 y) max (Da x Db y) ?
1 Da2 Db2 ? 1 - (x, y) ? 1 - (a0 x b0 y)
15- Good news
- Can handle constraints of this type
- (x, y) ? 1 - (1 x 1 y)
- easily (the so-called second-order conic
- programming (SOCP))
- Not much harder than linear programming!
16General Robust LP formulation
- Robust LP
-
- max cTx
- s/t A(i) x ? bi, i 1,,m
- where
- c, x ÃŽ Rn, A(i) ÃŽ R1 x n, A(i)A(i)0 wi Pi
- with
- wi ÃŽ R1 x ki, wi ? 1, i1,,m, Pi ÃŽ Rki x n
17- SOCP equivalent
- max cTx
- s/t Pi x ? bi - A(i)0 x, i 1,,m
- Probabilistic interpretation
- think of A(i) taken from an a-level set of your
favorite probability distribution (e.g.
multivariate normal) - the robust constraint will read
- satisfy the constraint with a given probability a
18Whered the ellipse come from?
- Expert opinion
- Statistics Averages live in ellipsoids
- Doesnt have to be an ellipse. Can be some other
shape (e.g., boxes)
19Software
- Commercial
- Mosek (http//www.mosek.com/)
- Free
- SeDuMi (http//sedumi.mcmaster.ca/)
- SDPT3.x (http//www.math.nus.edu.sg/mattohkc/sdpt
3.html/)
20Practicalities
- Realistic problem sizes
- number of variables/constraints on the order of
103 104 - depends (greatly) on the problem data
structure/sparsity - Possible to obtain a good, inexpensive
approximation with LP
21- Generality
- Possible to extend this approach to quite a few
other convex programming problems - Resources
- Lectures on Modern Convex Optimization Analysis,
Algorithms, and Engineering Applications by A.
Ben-Tal, A. S. Nemirovskii - Google for Robust Optimization (robust LP etc.)
22 Joint work with Millie Chu (Cornell) and Michael
B. Sharpe (Princess Margaret Hospital, Toronto)
23Cancer treatment
- About 1.3 million new cancer cases in the U.S.
each year - Nearly 60 receive radiation therapy (in
conjunction with surgery, chemotherapy etc)
24External beam radiation therapy
- Radiation delivered by a linear accelerator
- Cancer cells more susceptible than normal cells
- Overlay beams from different angles
- Dose given in daily fractions for 6 weeks
25Intensity Modulated Radiation Therapy
- Block parts of the radiation beam discretize
the whole beam into a grid of smaller beamlets - Choose different intensities for each beamlet
Intensity Modulated Radiation Therapy
Collaborative Working Group, 2001
26Treatment Planning
Goal Choose beam angles and beamlet
intensities that deliver enough radiation to kill
all tumor cells, while avoiding healthy organs
tissue as much as possible
- Take CT scan
- Delineate target region and healthy structures
- Discretize body as small cubes, or voxels
- Formulate solve a mathematical program to find
a good plan -
-
Princess Margaret Hospital
27Robust Treatment Planning
- Setup errors
- Patient motion
- Structural changes during treatment
- uncertainty in geometry
- Dont rescan patient much if at all
- Use RO to robustify mathematical program
28Model Formulation
Many different formulations exist we use a
penalty formulation minimize penalty
objective subject to Pr(Dose to voxel i in
healthy structure k Uk) 0.95 Pr(Dose to
voxel i in tumor L) 0.95 x
beamlet intensities 0
29Computational Results
- Prostate tumor 5 healthy regions
- 5 equi-spaced beams, 225 beamlets from each
angle - Voxel size 2 cm, 400 total voxels
- Solver Mosek, v. 3.0.1.18
- Solve time 6 seconds (LP), 45 minutes (SOCP)
30Dose-Volume Histograms
of structure receiving x Gy
deterministic solutions plan
DVH of expected dose
stochastic solutions plan
31Comparison
- Simulate 10 treatments (45 fractions each)
- For each of the 10 treatments, and for each
solution (deterministic stochastic), - calculated dose delivered to each voxel in each
fraction - summed over the 45 fractions to get total dose
delivered to each voxel - plotted DVH
32DVH Treatment 1
det
stoch
33DVH Treatment 2
det
stoch
34DVH Treatment 3
det
stoch
35DVH Treatment 4
det
stoch
36DVH Treatment 5
det
stoch
37DVH Treatment 6
det
stoch
38DVH Treatment 7
det
stoch
39DVH Treatment 8
det
stoch
40DVH Treatment 9
det
stoch
41DVH Treatment 10
det
stoch
42Conclusions
- LP pushes you into a corner
- True situation never same as data
- Robust LP Find good solution that is always
feasible within reason - Efficient solution methods can solve real
problems - Software available
43A Bit More Detail
- Di(x) Dose delivered to voxel i in N
fractions, with intensities x, a random variable - Di(x) is the sum of N random variables (N
45), assume iid, - apply CLT, so Di(x) is approximately
normally distributed - Take n sample shifts, s1,...,sn, with
associated probabilities p (p1,...,pn)T - Let ai()T ai(s1)T
- ai(s2)T dose delivered to voxel i,
shifted by sj, - from each beamlet with unit intensity
- ai(sn)T
- so that NpTai()Tx expected total dose
delivered to voxel i, for N fractions. - Let vi(x) sample variance of dose delivered
to voxel i - ? Di(x) Normal ( NpTai()Tx, Nvi(x) )
44Probabilistic Constraints
- Want constraints to be violated with low
probability (say, d .05) - Example maximum dose constraint on voxel i in
Hk - Assuming Di(x) Normal ( NpTai()Tx, Nvi(x) ),
?
?
mk
Want P(Di(x) gt mk) d
?
?
?
?
- Second order cone program (SOCP)
45Dose-Volume Constraints
- Physicians like constraints of formlt fraction
fk of structure Hk gets gt dk - 0-1 var for each voxel 1 if dose is gt dk.
- MIP Hard to solve!
- Many voxels get near max allowed dose
- Alternative upper bound the excess dose. For
healthy structure Hk, we require -
- Linear constraints?