SlideShare a Scribd company logo
1 of 28
Download to read offline
Duality Theory for Composite
Geometric Programming
Ya-Ping Wang
Department of Industrial Engineering
University of Pittsburgh
December 14, 2012
Duality	Theory	for	Composite	Geometric	Programming	
Page 1 of 27
Geometric Program (GP)
Primal posynomial program (GP)
(GP) 0inf ( ) . . ( ) 1, : {1,..., }m k
R
G s t G k K p

  
0<t
t t
Note that there are
 m design variables, 1
( , , ) , :{1,..., }mt t J m  t 0 ,
 each Gk(t) is a sum of terms indexed by the set [k]:
[ ]
( ): ( ), 0,1, , ,k ii k
G U k p
  t t
Across the p+1 different functions there are a total of
 n terms 1
, , , {1, , };whereU U I nn  
, ( ) : , 0,ija
i i j i ijj J
i I U C t C a R
    t ,
These terms are sequentially distributed into the (p+1)
problem functions as follows:
0 1[0] [1] [ ]; , 0< |[ ]|p kI p n n n n n k       
The “exponent matrix” is given by ,
, 1:[ ]n m
ij i ja A
Duality	Theory	for	Composite	Geometric	Programming	
Page 2 of 27
Equivalent Formulations of (GP)
In the variables = ln , ,j jz t j J
 ln ( ) ln ln , , lni i ij j i i ij J
U C a t c i I c C
       i
t a z
  [ ]
ˆln ( ) ln exp[ ] : ( ), .i
k i ki k
G c g k K
   t a z z
A convex formulation of GP:
(GP)z: 0inf ( ) . . ( ) 0,m k
R
g s t g k K

  
z
z z
Duality	Theory	for	Composite	Geometric	Programming	
Page 3 of 27
Equivalent Formulations of (GP) (cont’d)
In the variables =:ix i Ii
a z,
 [ ]
( ) ln ( ) ln exp( ) : ( )k k
k k i ii k
g G x c geo
    z t x c
where [ ]
( ) : ln[ exp ] : knk
ii k
geo x R R
 x is called a
geometric function (also called logexp(x)), [ ][ ]k
i i kx x
and [ ][ ]k
i i kc c .
A GGP (Generalized Geometric Programming)
formulation of GP: (GP)x
0 0
inf ( ) . . ( ) 0, ,k k
geo s t geo k K     x x c x c x P
where  | m n
R R  Az zP is the column space of A.
Duality	Theory	for	Composite	Geometric	Programming	
Page 4 of 27
Dual Posynomial Program (GD)
 ˆ [ ]
0
1
[ ]
sup ( ) : (Dual function)
. . 0, , 1 (normality condition)
0, (orthogonality conditions)
where : ,
i
n
i k ik K i k
R
i
n
ij i
i
k i
i k
V C
s t i I
a j J

 
 

 
 




   
  
 
 




ˆ.k K







 

Each dual variable i corresponds to a primal term iU .
Degree of difficulty is defined as ( 1)d n m  
The log-dual function
 
0 [ ]
1 1
( ) :ln ( ) ln /
( ln ) ln
p
i i k i
k i k
pn
i i i k k
i k
v V C
c
  
   
 
 
    
  

 
 
is concave on its domain of definition, and differentiable
in the interior of its domain.
Duality	Theory	for	Composite	Geometric	Programming	
Page 5 of 27
Main Lemma of GP
Lemma: If t is feasible for primal program (GP) and  is
feasible for dual program (GD), then
0 ( ) ( )G Vt  .
Moreover, under the same conditions 0 ( ) ( )G Vt  if, and
only if, one of the following two sets of equivalent
extremality conditions holds:
I.
( ) 1, , (1)
ˆ( ) ( ), [ ], (2)
k
k
i k k i
G k K
G U i k k K

 
   

    
t
t t
II.
0( ) ( ), [0] (1)
( ), [ ], (2)
i
i
k i
U G i
U i k k K



 
  
t t
t
in which case t is optimal for primal program (GP) and 
is optimal for dual program (GD).
Duality	Theory	for	Composite	Geometric	Programming	
Page 6 of 27
Dual to Primal Conversion at Optimality
When 0 ( ) ( )G Vt  ,
we get from condition II on p.5 the equations
( ), [0]
( )
, [ ], , and 0
i
i
i k k
V i
U
i k k K

  
 
 
  
t

,
which are log-linear:
 1
ln( ( )), [0]
ln , [ ], for which 0
m
i
ij j i
j i k k
V i
a z c
i k k K

  

  
  


So, knowing the optimal dual solution  an optimal
primal solution can in general be easily recovered by
solving this log-linear system for z and hence for t.
Duality	Theory	for	Composite	Geometric	Programming	
Page 7 of 27
EXAMPLE:
(Maximum Likelihood Estimator of a Bernoulli Parameter)
Suppose that n independent trials, each of which is a
success with probability p, are performed. What is the
maximum likelihood estimator (MLE) of p?
The likelihood function to be maximized is
1, 1
( , | ) (1 ) , with , 0 or 1,n n
ns n s
n n i if x x p p p s x x i
    
Example: n=3, m=2: 1
( , ) 0
inf s.t. 1n ns s n
p q
f p q p q 

  
The exponent matrix and a unique solution from
extremality condition II (2) on p.5:
* *
1
1 *
*2
2 *
1
3
=
1
ˆ1 0 , so
0 1
n n
n n
n
n
p q n
s s n
s S
s p p X
n n
n s
 





  
      
   
Duality	Theory	for	Composite	Geometric	Programming	
Page 8 of 27
Prior Extensions to GP
 Signomial GP which allow some <0iC  non-
convex programs: duality results are not strong
 Peterson’s Generalized GP (GGP)More general
separable convex programs
Extensions to GP investigated in this thesis
 Composite GP (CGP): itself a special case of
Peterson’s GGP. It includes as special cases:
 Exponential GP (EGP)
 Quadratic GP (QGP)
 (lpGP)
We start with Exponential GP (EGP) using a motivating
example…
Duality	Theory	for	Composite	Geometric	Programming	
Page 9 of 27
A Motivational Example
(Maximum Likelihood Estimator of a Poisson Parameter)
Suppose 1, , nX X are independent Poisson random
variables each having mean λ. Determine the MLE of λ.
The likelihood function to be maximized is
1, 1 1
( , | ) / ( ! !), where :n
ns n
n n n ii
f x x e x x s x
  

    
Equivalently, one can 0min ns n
e 
 
  . Although this is
not a posynomial, it can be solved as an EGP:
1
1 1
1
From condition (e) on p.14, we get
1 ˆ/ = , so /
1
n
n n
n
y s
y s n S n X
s

  

 
   
 
This same method also works for Exponential and Normal
parameters…
Duality	Theory	for	Composite	Geometric	Programming	
Page 10 of 27
Exponential Geometric Program (EGP)
In the previous example:
 a posynomial term is multiplied by an exponential
factor of another posynomial term.
This gives rise to an EGP problem where some
posynomial term ( )iU t is multiplied by an exponential
factor of another posynomial ( ): ( )i ll i
E V 
 t t .
Primal EGP problem (EGP): (cf (GP) on p.1)
0inf ( ) . . ( ) 1, ,m kR
G s t G k K
 0
 
<t
t t
where
[ ]
1
ˆ( ) : ( ) exp ( ) , : {0}
( ) , ,and ( ) ,
{1, , } 1 ,| | 0,
ij lj
k i li k l i
a b
i i j l l jj J j J
i n
G U V k K K
U C t i I V D t l L
L r n i r r r r
  
 
      
     
            
 
 
t t t
t t

 
[ ]:ljB b r m 
Duality	Theory	for	Composite	Geometric	Programming	
Page 11 of 27
Equivalent formulations of (EGP)
In the variables
A convex form of (EGP): (EGP)z
0inf ( ) . . ( ) 0,m k
R
g s t g k K

   
z
z z , where
  [ ]
ˆ( ) :ln ( ) ln exp exp ,i
k k i li k l i
g G c d k K  
        l
z t a z b z +
In the variables : , : ,i lx i I l L   i l
a z, b z,
[ ] [ ]
( ) ( , ) with exp,k k k k
k lg geo h l L      z x c ξ d
where
 [ ]
[ ]
( , ): ln exp exp( )k k
i li k l i
geo x   
    x ξ
:
k
kn r
R R R  is called an exponential geometric
function, [ ]
[ ] [ ]
[ ] , [ ] , :irk i i k
i k l l i ii k
R r r   
    ξ ξ ξ ,
and [ ]k
d is similarly defined.
= ln , ,j jz t j J
Duality	Theory	for	Composite	Geometric	Programming	
Page 12 of 27
Equivalent formulations of (EGP) (cont’d)
A GGP formulation of (EGP):
(EGP)x,ξ
0 0 [0] [0]
( )
[ ] [ ]
inf ( , )
. .
( , ) 0, ,( )
n r
R R
k k k k
geo
s t
geo k K
 
  


      
x,
x c ξ d
x c ξ d x,ξ



P
where  | m n r
M R R 
  z zP is the column space of the
composite exponent matrix
A
B
M
 
  
 
.
Duality	Theory	for	Composite	Geometric	Programming	
Page 13 of 27
Dual of EGP program (EGD) (cf (GD) on p.4)
ˆ( ) ( , ) [ ]
0
1 1
sup ( , ) :
. . 0, , 0, , 1,
0, (Orthogonality Conditions)
0
i l
n r
y
i k l i
R R i k l ik K i Ii l
i l
n r
ij i lj l
i l
i
C D y
V e
y
s t y i I l L
a y b j J
y



 


    
 
   
     
   
      
  
 
   
 
,
y
y 

[ ]
0, , (*) p.16
ˆwhere : , , : =
l
k i l
i k l L
l i i I
y k K  

 










       

   

 
Note that ( ), ( )i i l ly U V  t t
and degree of difficulty ( 1)d n r m    →p.4
The log-dual function (cf p.4)
   
ˆ [ ]
( , ) ln / ln /i i k i l l i l
i k l ik K i I
v y C y D y   
   
       y 
is usc proper concave in( , )y  , and differentiable in the
interior of its domain.
Duality	Theory	for	Composite	Geometric	Programming	
Page 14 of 27
Main Lemma of EGP
Lemma If t is feasible for primal program (EGP) and
( , )y  is feasible for dual program (EGD), then 0 ( ) ( , )G V y t 
Moreover, under the same conditions, 0 ( ) ( , )G V y t  if
and only if one of the following two sets of equivalent
extremality conditions hold: (cf		I	&	II	on	p.5)
I' Condition (e) & I
( ) 1, ,
ˆ( ) ( ), [ ],
k
k
i k k i
G k K
y G U i k k K


  

  

 
t
t t
II' Condition (e) & II
0( ) ( ), [0]
( ), [ ],
i
i
k i
U G i
y
U i k k K
 
 
 


t t
t
where (e) ( ), ,l i lyV l i i I 
   t
in which case t is optimal for primal program (EGP) and
( , )y  is optimal for dual program (EGD).
	p.9 Solve the previous motivational Example.
Duality	Theory	for	Composite	Geometric	Programming	
Page 15 of 27
Composite Geometric Program (CGP)
Primal CGP Program (CGP) (cf (EGP) on p.10)
0inf ( ) . . ( ) 1,m kR
G s t G k K
 0
 
<t
t t , where
[ ]
ˆ( ) ( ) exp (ln ( )) ,k i l li k l i
G U h V k K  
       t t t ,
:lh R R is a differentiable and strictly convex function.
Note that:
(QGP): if 21
2( ) ,lh l L   
(lpGP): if ( ) | | / , where 1,lp
l l lh p p l L    
Duality	Theory	for	Composite	Geometric	Programming	
Page 16 of 27
Composite Geometric Program (CGP) (cont’d)
A Convex Form of (CGP): (CGP)z
0inf ( ) . . ( ) 0,m kR
g s t g k K
   z
z z , where
  [ ]
( ): ln ( ) ln exp i l
k k i l li k l i
g G c h d  
        z t a z b z
A GGP Form of (CGP): (CGP)x, 		
0 [0] [ ]
0
( )
inf ( , ) . . ( , ) 0, , ( )n r
k k
k
R R
f s t f k K
 
   ξ ξ ξ
x,
x x x, P

with   [ ]
( , ) ( , ) : ln expk i l li k l i
f geo x h   
          
Duality	Theory	for	Composite	Geometric	Programming	
Page 17 of 27
Dual program (CGD) (cf (EGD) on p.13)
*
ˆ( ) [ ]
0
1 1
sup ( , ): exp ( / )
. . 0, , 1, (Normality Condition)
0, (Ortho
i
n r
y
i k
l l i l l i
R R l ii kk K i Ii
i
n r
ij i lj l
i l
C
V d y h y
y
s t y i I
a y b j J

 


    
 
   
       
  
   
  
 
 
y,η
y 
[ ]
gonality Conditions)
, , (**)
ˆwhere : , .
l i l
k i
i k
y J l i i I
y k K














     

   


where *
lh , the conjugate of lh , has domain interval lJ .
(**) 0l  in (EGD) on p. 13 and that in any (CGD)
(*) 0 0, ,i ly l i i I 
       
The log-dual function (cf p.13)
  *
ˆ [ ]
( , ) ln / ( / )i i k i l l i l l i
i k l ik K i I
v y C y d y h y  
   
          y 
is usc proper concave in( , )y  , and differentiable in the
interior of its domain.
Duality	Theory	for	Composite	Geometric	Programming	
Page 18 of 27
Main Lemma of CGP
Lemma: If t is feasible for primal program (CGP) and
( , )y  is feasible for dual program (CGD), then
0 ( ) ( , )G V t y  .
Moreover, under the same conditions, 0 ( ) ( , )G V t y  if
and only if one of the following two sets of equivalent
extremality conditions hold: (cf		I’	&	II’	on	p.14)
I'' Condition (c) & I
( ) 1, ,
ˆ( ) ( ), [ ],
k
k
i k k i
G k K
y G U i k k K


  

  

 
t
t t
II'' Condition (c) & II
0( ) ( ), [0]
( ), [ ],
i
i
k i
U G i
y
U i k k K
 
 
 


t t
t
where (c) '
(ln ( )), ,ll i ly h V l i i I 
   t
in which case t is optimal for primal program (CGP) and
( , )y  is optimal for dual program (CGD).
(lpGP): (c) 1
| ln ( ) | sgn(ln ( )), ,lp
l i l ly V V l i i I  
   t t
Duality	Theory	for	Composite	Geometric	Programming	
Page 19 of 27
First Duality Theorem of CGP
NOTE: The statements for EGP, QGP, lpGP cases are
almost the same.
Theorem: Suppose that primal program (CGP) is super-
consistent. Then the following three conditions are
equivalent:
1) t’ is a minimal solution to (CGP).
2) There exists a vector ' p
R for z′ (where z′= ln t′)
such that ( ', ')z  forms a saddle point of ( , )l z  .
where ( , )l z  is the Lagrangian of (CGP)z.
Duality	Theory	for	Composite	Geometric	Programming	
Page 20 of 27
3) There exists a vector ' p
R for z′ (where z′= ln t′)
such that ( ', ')z  satisfies the KKT conditions for
(CGP)z:
in which case the set of all such vectors 'λ is a non-empty
compact convex subset of p
R , and the dual program
(CGD) also has a maximum solution ( ', ')y  such that
0min(CGP) ( ') ( ', ') max(CGD)G V   t y 
and they satisfy the extremality conditions I''& II'' on
p.18.
(Perfect Duality)
' '
' '
0ˆ
( ) 0, ( ) 0, ( ) 0,
( ) ( , ') ( ) , 1
k k k k
z k kk K
a g g k K
b l g where
 
 
    
     0
 
 
z' z'
z' z'
Duality	Theory	for	Composite	Geometric	Programming	
Page 21 of 27
Second Duality Theorem of EGP
From linear programming duality theory:
Of the following two linear systems, exactly one has a
solution (where  :
A
B
M n r m
 
  
 
):
(I) Find z with 0 0
 
  
 
A
B
z
(II) Find >0 with T T
A B
 
 
 
 
y
y 0

We say that program (EGD) is canonical if system (II)
has a solution when M is the program’s composite
exponent matrix.
Theorem: Suppose that primal program (EGP) is
consistent. Then the minimum set of program (EGP)z is
non-empty and bounded if, and only if, dual program
(EGD) is canonical, in which case program (EGP) has a
minimum solution t′.
Duality	Theory	for	Composite	Geometric	Programming	
Page 22 of 27
Second Duality Theorem of lpGP*
* This theorem also applies to its special case QGP.
Again, from linear programming duality theory
Of the following two linear systems exactly one has a
solution (where  :
A
B
M n r m
 
   
 
):
(I) Find z with ,  0 Az 0 Bz 0
(II) Find with , and T T
A B
 
 
 
  
y
y 0 y 0


We say that program (lpGD) is canonical if system (II)
has a solution when M is the program’s composite
exponent matrix.
Theorem 7.3.2 Suppose that primal program (lpGP) is
consistent. Then the minimum set of program (lpGP)z is
non-empty and bounded if and only if its dual program
(lpGD) is canonical.
Duality	Theory	for	Composite	Geometric	Programming	
Page 23 of 27
SUMMARY OF CONTRIBUTIONS
 Extension of traditional GP models to
o to the more general Exponential GP models,
o to the even more general Composite GP models,
which include as important special cases:
 (EGP)
 (lpGP)
 (QGP)
 Showing that all of these are special cases of
Peterson’s GGP models (for which he has given a
Main Lemma but not a First or a Second duality
theorem).
Duality	Theory	for	Composite	Geometric	Programming	
Page 24 of 27
CONTRIBUTIONS (cont’d)
 For each extension:
o multiple direct proofs of the Main Lemma, with a
2nd set of equivalent Extremality Conditions.
 For the First Duality Theorem, proof that
o a superconsistent primal program (CGP) has a
minimal solution 't if, and only if, there exists a
vector ' p
R such that ( ', ')z  forms a saddle
point of the Lagrangian ( , )l z  and if, and only
if, it satisfies the KKT conditions for (CGP)z,
o in which case the set of all such Lagrange
multiplier vectors λ is a non-empty compact
convex subset of p
R , whereas the original
theorem is a “If…then” statement and only
showed the existence ofλ for the (GP) case only.
Duality	Theory	for	Composite	Geometric	Programming	
Page 25 of 27
 For the Second Duality Theorem, proof that
o for the special (and more important) cases of
(QGP), (lpGP) and (EGP),
 the minimum set of a consistent primal
program is non-empty and bounded if and
only if its dual program is canonical
 the original theorem showed the existence of
a primal solution for the (GP) case only,
when its dual program is canonical.
 Sensitivity Analysis: If *
z is optimal for (CGP)z:
* * *
* * *0 0 0
* *
* * * *0 0
( ) ( ) ( )
, , ,
( ) ( )
, , where ln
i l k
i l k
i j l j k k
ij lj
g g g
y
c d b
g g
y z z b B
a b
 

  
   
  
 
  
 
  
 
z z z
z z
 Dual to primal conversion of optimal solutions
when there is no duality gap: 0 ( ) ( , )G V y t  .
Duality	Theory	for	Composite	Geometric	Programming	
Page 26 of 27
Directions for Future Research
1.Development of proofs of a strong version of the
First Duality Theorem of (CGP): If a primal program
(CGP) is superconsistent and has a finite infimum,
then the dual program (CGD) has a maximum
solution ( ', ')y  such that
inf(CGP) max(CGD) ( ', ')V  y 
2.Development of computational algorithms.
Duality	Theory	for	Composite	Geometric	Programming	
Page 27 of 27
QUESTIONS

More Related Content

What's hot

R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...
R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...
R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...Matt Moores
 
A common random fixed point theorem for rational inequality in hilbert space
A common random fixed point theorem for rational inequality in hilbert spaceA common random fixed point theorem for rational inequality in hilbert space
A common random fixed point theorem for rational inequality in hilbert spaceAlexander Decker
 
A Note on Latent LSTM Allocation
A Note on Latent LSTM AllocationA Note on Latent LSTM Allocation
A Note on Latent LSTM AllocationTomonari Masada
 
Analysing and combining partial problem solutions for properly informed heuri...
Analysing and combining partial problem solutions for properly informed heuri...Analysing and combining partial problem solutions for properly informed heuri...
Analysing and combining partial problem solutions for properly informed heuri...Alexander Decker
 
Accelerating Pseudo-Marginal MCMC using Gaussian Processes
Accelerating Pseudo-Marginal MCMC using Gaussian ProcessesAccelerating Pseudo-Marginal MCMC using Gaussian Processes
Accelerating Pseudo-Marginal MCMC using Gaussian ProcessesMatt Moores
 
Multilayer Neural Networks
Multilayer Neural NetworksMultilayer Neural Networks
Multilayer Neural NetworksESCOM
 
Modification of a heuristic method
Modification of a heuristic methodModification of a heuristic method
Modification of a heuristic methodorajjournal
 
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...Matt Moores
 
Gentle Introduction to Dirichlet Processes
Gentle Introduction to Dirichlet ProcessesGentle Introduction to Dirichlet Processes
Gentle Introduction to Dirichlet ProcessesYap Wooi Hen
 
Low-rank tensor approximation (Introduction)
Low-rank tensor approximation (Introduction)Low-rank tensor approximation (Introduction)
Low-rank tensor approximation (Introduction)Alexander Litvinenko
 
ICML2013読み会 Large-Scale Learning with Less RAM via Randomization
ICML2013読み会 Large-Scale Learning with Less RAM via RandomizationICML2013読み会 Large-Scale Learning with Less RAM via Randomization
ICML2013読み会 Large-Scale Learning with Less RAM via RandomizationHidekazu Oiwa
 
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...garbervetsky
 
On the equality of the grundy numbers of a graph
On the equality of the grundy numbers of a graphOn the equality of the grundy numbers of a graph
On the equality of the grundy numbers of a graphijngnjournal
 
Graph Kernels for Chemical Informatics
Graph Kernels for Chemical InformaticsGraph Kernels for Chemical Informatics
Graph Kernels for Chemical InformaticsMukund Raj
 
A common unique random fixed point theorem in hilbert space using integral ty...
A common unique random fixed point theorem in hilbert space using integral ty...A common unique random fixed point theorem in hilbert space using integral ty...
A common unique random fixed point theorem in hilbert space using integral ty...Alexander Decker
 

What's hot (19)

R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...
R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...
R package 'bayesImageS': a case study in Bayesian computation using Rcpp and ...
 
A common random fixed point theorem for rational inequality in hilbert space
A common random fixed point theorem for rational inequality in hilbert spaceA common random fixed point theorem for rational inequality in hilbert space
A common random fixed point theorem for rational inequality in hilbert space
 
A Note on Latent LSTM Allocation
A Note on Latent LSTM AllocationA Note on Latent LSTM Allocation
A Note on Latent LSTM Allocation
 
Analysing and combining partial problem solutions for properly informed heuri...
Analysing and combining partial problem solutions for properly informed heuri...Analysing and combining partial problem solutions for properly informed heuri...
Analysing and combining partial problem solutions for properly informed heuri...
 
Accelerating Pseudo-Marginal MCMC using Gaussian Processes
Accelerating Pseudo-Marginal MCMC using Gaussian ProcessesAccelerating Pseudo-Marginal MCMC using Gaussian Processes
Accelerating Pseudo-Marginal MCMC using Gaussian Processes
 
Multilayer Neural Networks
Multilayer Neural NetworksMultilayer Neural Networks
Multilayer Neural Networks
 
Modification of a heuristic method
Modification of a heuristic methodModification of a heuristic method
Modification of a heuristic method
 
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...
bayesImageS: Bayesian computation for medical Image Segmentation using a hidd...
 
Gentle Introduction to Dirichlet Processes
Gentle Introduction to Dirichlet ProcessesGentle Introduction to Dirichlet Processes
Gentle Introduction to Dirichlet Processes
 
Ceske budevice
Ceske budeviceCeske budevice
Ceske budevice
 
Lecture12 xing
Lecture12 xingLecture12 xing
Lecture12 xing
 
Low-rank tensor approximation (Introduction)
Low-rank tensor approximation (Introduction)Low-rank tensor approximation (Introduction)
Low-rank tensor approximation (Introduction)
 
Triggering patterns of topology changes in dynamic attributed graphs
Triggering patterns of topology changes in dynamic attributed graphsTriggering patterns of topology changes in dynamic attributed graphs
Triggering patterns of topology changes in dynamic attributed graphs
 
ICML2013読み会 Large-Scale Learning with Less RAM via Randomization
ICML2013読み会 Large-Scale Learning with Less RAM via RandomizationICML2013読み会 Large-Scale Learning with Less RAM via Randomization
ICML2013読み会 Large-Scale Learning with Less RAM via Randomization
 
20 sameer sharma 186-206
20 sameer sharma 186-20620 sameer sharma 186-206
20 sameer sharma 186-206
 
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...
ByteCode 2012 Talk: Quantitative analysis of Java/.Net like programs to under...
 
On the equality of the grundy numbers of a graph
On the equality of the grundy numbers of a graphOn the equality of the grundy numbers of a graph
On the equality of the grundy numbers of a graph
 
Graph Kernels for Chemical Informatics
Graph Kernels for Chemical InformaticsGraph Kernels for Chemical Informatics
Graph Kernels for Chemical Informatics
 
A common unique random fixed point theorem in hilbert space using integral ty...
A common unique random fixed point theorem in hilbert space using integral ty...A common unique random fixed point theorem in hilbert space using integral ty...
A common unique random fixed point theorem in hilbert space using integral ty...
 

Viewers also liked

Informática y convergencia tecnologica
Informática y convergencia tecnologicaInformática y convergencia tecnologica
Informática y convergencia tecnologicaPaola Rodríguez
 
Tunnukset ja Yksiköt
Tunnukset ja YksikötTunnukset ja Yksiköt
Tunnukset ja YksikötSuvi Muotka
 
Doenças sexualmente transmissíveis nathalia[
Doenças sexualmente transmissíveis nathalia[Doenças sexualmente transmissíveis nathalia[
Doenças sexualmente transmissíveis nathalia[NathNascimento
 
(DL)Tapes and Bandage Catalogue
(DL)Tapes and Bandage Catalogue(DL)Tapes and Bandage Catalogue
(DL)Tapes and Bandage CatalogueKintape DL
 
entregable primera parte
entregable primera parte entregable primera parte
entregable primera parte ana velez
 
Perspectivas de la comunicación
Perspectivas de la comunicaciónPerspectivas de la comunicación
Perspectivas de la comunicaciónMaria Camila
 
Tecnología de la Educación
Tecnología de la Educación Tecnología de la Educación
Tecnología de la Educación masielyadira
 

Viewers also liked (10)

Informática y convergencia tecnologica
Informática y convergencia tecnologicaInformática y convergencia tecnologica
Informática y convergencia tecnologica
 
Tunnukset ja Yksiköt
Tunnukset ja YksikötTunnukset ja Yksiköt
Tunnukset ja Yksiköt
 
News agencies
News agenciesNews agencies
News agencies
 
Doenças sexualmente transmissíveis nathalia[
Doenças sexualmente transmissíveis nathalia[Doenças sexualmente transmissíveis nathalia[
Doenças sexualmente transmissíveis nathalia[
 
(DL)Tapes and Bandage Catalogue
(DL)Tapes and Bandage Catalogue(DL)Tapes and Bandage Catalogue
(DL)Tapes and Bandage Catalogue
 
El reportaje1
El reportaje1El reportaje1
El reportaje1
 
entregable primera parte
entregable primera parte entregable primera parte
entregable primera parte
 
Google drive
Google driveGoogle drive
Google drive
 
Perspectivas de la comunicación
Perspectivas de la comunicaciónPerspectivas de la comunicación
Perspectivas de la comunicación
 
Tecnología de la Educación
Tecnología de la Educación Tecnología de la Educación
Tecnología de la Educación
 

Similar to MLE Estimation via EGP Duality

Double Robustness: Theory and Applications with Missing Data
Double Robustness: Theory and Applications with Missing DataDouble Robustness: Theory and Applications with Missing Data
Double Robustness: Theory and Applications with Missing DataLu Mao
 
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMS
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMSSOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMS
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMSTahia ZERIZER
 
MAPE regression, seminar @ QUT (Brisbane)
MAPE regression, seminar @ QUT (Brisbane)MAPE regression, seminar @ QUT (Brisbane)
MAPE regression, seminar @ QUT (Brisbane)Arnaud de Myttenaere
 
D242228
D242228D242228
D242228irjes
 
International Refereed Journal of Engineering and Science (IRJES)
International Refereed Journal of Engineering and Science (IRJES)International Refereed Journal of Engineering and Science (IRJES)
International Refereed Journal of Engineering and Science (IRJES)irjes
 
Optimal control problem for processes
Optimal control problem for processesOptimal control problem for processes
Optimal control problem for processesIJCI JOURNAL
 
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...Alexander Decker
 
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...Shu Tanaka
 
Complexity Classes and the Graph Isomorphism Problem
Complexity Classes and the Graph Isomorphism ProblemComplexity Classes and the Graph Isomorphism Problem
Complexity Classes and the Graph Isomorphism Problemcseiitgn
 
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtures
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: MixturesCVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtures
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtureszukun
 
On the-approximate-solution-of-a-nonlinear-singular-integral-equation
On the-approximate-solution-of-a-nonlinear-singular-integral-equationOn the-approximate-solution-of-a-nonlinear-singular-integral-equation
On the-approximate-solution-of-a-nonlinear-singular-integral-equationCemal Ardil
 
Some Fixed Point Theorems in b G -cone Metric Space
Some Fixed Point Theorems in b G -cone Metric Space Some Fixed Point Theorems in b G -cone Metric Space
Some Fixed Point Theorems in b G -cone Metric Space Komal Goyal
 
A common fixed point of integral type contraction in generalized metric spacess
A  common fixed point of integral type contraction in generalized metric spacessA  common fixed point of integral type contraction in generalized metric spacess
A common fixed point of integral type contraction in generalized metric spacessAlexander Decker
 
Lecture 2 data structures and algorithms
Lecture 2 data structures and algorithmsLecture 2 data structures and algorithms
Lecture 2 data structures and algorithmsAakash deep Singhal
 
An Approach For Solving Nonlinear Programming Problems
An Approach For Solving Nonlinear Programming ProblemsAn Approach For Solving Nonlinear Programming Problems
An Approach For Solving Nonlinear Programming ProblemsMary Montoya
 

Similar to MLE Estimation via EGP Duality (20)

Double Robustness: Theory and Applications with Missing Data
Double Robustness: Theory and Applications with Missing DataDouble Robustness: Theory and Applications with Missing Data
Double Robustness: Theory and Applications with Missing Data
 
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMS
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMSSOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMS
SOLVING BVPs OF SINGULARLY PERTURBED DISCRETE SYSTEMS
 
MAPE regression, seminar @ QUT (Brisbane)
MAPE regression, seminar @ QUT (Brisbane)MAPE regression, seminar @ QUT (Brisbane)
MAPE regression, seminar @ QUT (Brisbane)
 
D242228
D242228D242228
D242228
 
International Refereed Journal of Engineering and Science (IRJES)
International Refereed Journal of Engineering and Science (IRJES)International Refereed Journal of Engineering and Science (IRJES)
International Refereed Journal of Engineering and Science (IRJES)
 
Optimal control problem for processes
Optimal control problem for processesOptimal control problem for processes
Optimal control problem for processes
 
MUMS: Bayesian, Fiducial, and Frequentist Conference - Model Selection in the...
MUMS: Bayesian, Fiducial, and Frequentist Conference - Model Selection in the...MUMS: Bayesian, Fiducial, and Frequentist Conference - Model Selection in the...
MUMS: Bayesian, Fiducial, and Frequentist Conference - Model Selection in the...
 
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...
1.[1 5]implementation of pre compensation fuzzy for a cascade pid controller ...
 
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...
Quantum Annealing for Dirichlet Process Mixture Models with Applications to N...
 
Complexity Classes and the Graph Isomorphism Problem
Complexity Classes and the Graph Isomorphism ProblemComplexity Classes and the Graph Isomorphism Problem
Complexity Classes and the Graph Isomorphism Problem
 
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtures
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: MixturesCVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtures
CVPR2010: Advanced ITinCVPR in a Nutshell: part 6: Mixtures
 
On the-approximate-solution-of-a-nonlinear-singular-integral-equation
On the-approximate-solution-of-a-nonlinear-singular-integral-equationOn the-approximate-solution-of-a-nonlinear-singular-integral-equation
On the-approximate-solution-of-a-nonlinear-singular-integral-equation
 
Some Fixed Point Theorems in b G -cone Metric Space
Some Fixed Point Theorems in b G -cone Metric Space Some Fixed Point Theorems in b G -cone Metric Space
Some Fixed Point Theorems in b G -cone Metric Space
 
02 asymp
02 asymp02 asymp
02 asymp
 
A common fixed point of integral type contraction in generalized metric spacess
A  common fixed point of integral type contraction in generalized metric spacessA  common fixed point of integral type contraction in generalized metric spacess
A common fixed point of integral type contraction in generalized metric spacess
 
Slides_A4.pdf
Slides_A4.pdfSlides_A4.pdf
Slides_A4.pdf
 
Lecture 2 data structures and algorithms
Lecture 2 data structures and algorithmsLecture 2 data structures and algorithms
Lecture 2 data structures and algorithms
 
An Approach For Solving Nonlinear Programming Problems
An Approach For Solving Nonlinear Programming ProblemsAn Approach For Solving Nonlinear Programming Problems
An Approach For Solving Nonlinear Programming Problems
 
MUMS Opening Workshop - An Overview of Reduced-Order Models and Emulators (ED...
MUMS Opening Workshop - An Overview of Reduced-Order Models and Emulators (ED...MUMS Opening Workshop - An Overview of Reduced-Order Models and Emulators (ED...
MUMS Opening Workshop - An Overview of Reduced-Order Models and Emulators (ED...
 
Lect02.ppt
Lect02.pptLect02.ppt
Lect02.ppt
 

Recently uploaded

Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations120cr0395
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxupamatechverse
 
Coefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptxCoefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptxAsutosh Ranjan
 
UNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular ConduitsUNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular Conduitsrknatarajan
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)Suman Mia
 
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINEMANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINESIVASHANKAR N
 
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...ranjana rawat
 
result management system report for college project
result management system report for college projectresult management system report for college project
result management system report for college projectTonystark477637
 
Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingrakeshbaidya232001
 
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...ranjana rawat
 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...roncy bisnoi
 
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).pptssuser5c9d4b1
 
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service Nashik
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service NashikCall Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service Nashik
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service NashikCall Girls in Nagpur High Profile
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSRajkumarAkumalla
 
UNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its PerformanceUNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its Performancesivaprakash250
 
UNIT-III FMM. DIMENSIONAL ANALYSIS
UNIT-III FMM.        DIMENSIONAL ANALYSISUNIT-III FMM.        DIMENSIONAL ANALYSIS
UNIT-III FMM. DIMENSIONAL ANALYSISrknatarajan
 
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur High Profile
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...ranjana rawat
 

Recently uploaded (20)

Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur EscortsCall Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
Call Girls Service Nagpur Tanvi Call 7001035870 Meet With Nagpur Escorts
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptx
 
Coefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptxCoefficient of Thermal Expansion and their Importance.pptx
Coefficient of Thermal Expansion and their Importance.pptx
 
UNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular ConduitsUNIT-II FMM-Flow Through Circular Conduits
UNIT-II FMM-Flow Through Circular Conduits
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
 
Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024Water Industry Process Automation & Control Monthly - April 2024
Water Industry Process Automation & Control Monthly - April 2024
 
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINEMANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
MANUFACTURING PROCESS-II UNIT-2 LATHE MACHINE
 
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANJALI) Dange Chowk Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
 
result management system report for college project
result management system report for college projectresult management system report for college project
result management system report for college project
 
Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writing
 
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
The Most Attractive Pune Call Girls Manchar 8250192130 Will You Miss This Cha...
 
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
Call Girls Pimpri Chinchwad Call Me 7737669865 Budget Friendly No Advance Boo...
 
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt
247267395-1-Symmetric-and-distributed-shared-memory-architectures-ppt (1).ppt
 
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service Nashik
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service NashikCall Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service Nashik
Call Girls Service Nashik Vaishnavi 7001305949 Independent Escort Service Nashik
 
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICSHARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
HARDNESS, FRACTURE TOUGHNESS AND STRENGTH OF CERAMICS
 
UNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its PerformanceUNIT - IV - Air Compressors and its Performance
UNIT - IV - Air Compressors and its Performance
 
UNIT-III FMM. DIMENSIONAL ANALYSIS
UNIT-III FMM.        DIMENSIONAL ANALYSISUNIT-III FMM.        DIMENSIONAL ANALYSIS
UNIT-III FMM. DIMENSIONAL ANALYSIS
 
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur EscortsCall Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
Call Girls in Nagpur Suman Call 7001035870 Meet With Nagpur Escorts
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
 

MLE Estimation via EGP Duality

  • 1. Duality Theory for Composite Geometric Programming Ya-Ping Wang Department of Industrial Engineering University of Pittsburgh December 14, 2012
  • 2. Duality Theory for Composite Geometric Programming Page 1 of 27 Geometric Program (GP) Primal posynomial program (GP) (GP) 0inf ( ) . . ( ) 1, : {1,..., }m k R G s t G k K p     0<t t t Note that there are  m design variables, 1 ( , , ) , :{1,..., }mt t J m  t 0 ,  each Gk(t) is a sum of terms indexed by the set [k]: [ ] ( ): ( ), 0,1, , ,k ii k G U k p   t t Across the p+1 different functions there are a total of  n terms 1 , , , {1, , };whereU U I nn   , ( ) : , 0,ija i i j i ijj J i I U C t C a R     t , These terms are sequentially distributed into the (p+1) problem functions as follows: 0 1[0] [1] [ ]; , 0< |[ ]|p kI p n n n n n k        The “exponent matrix” is given by , , 1:[ ]n m ij i ja A
  • 3. Duality Theory for Composite Geometric Programming Page 2 of 27 Equivalent Formulations of (GP) In the variables = ln , ,j jz t j J  ln ( ) ln ln , , lni i ij j i i ij J U C a t c i I c C        i t a z   [ ] ˆln ( ) ln exp[ ] : ( ), .i k i ki k G c g k K    t a z z A convex formulation of GP: (GP)z: 0inf ( ) . . ( ) 0,m k R g s t g k K     z z z
  • 4. Duality Theory for Composite Geometric Programming Page 3 of 27 Equivalent Formulations of (GP) (cont’d) In the variables =:ix i Ii a z,  [ ] ( ) ln ( ) ln exp( ) : ( )k k k k i ii k g G x c geo     z t x c where [ ] ( ) : ln[ exp ] : knk ii k geo x R R  x is called a geometric function (also called logexp(x)), [ ][ ]k i i kx x and [ ][ ]k i i kc c . A GGP (Generalized Geometric Programming) formulation of GP: (GP)x 0 0 inf ( ) . . ( ) 0, ,k k geo s t geo k K     x x c x c x P where  | m n R R  Az zP is the column space of A.
  • 5. Duality Theory for Composite Geometric Programming Page 4 of 27 Dual Posynomial Program (GD)  ˆ [ ] 0 1 [ ] sup ( ) : (Dual function) . . 0, , 1 (normality condition) 0, (orthogonality conditions) where : , i n i k ik K i k R i n ij i i k i i k V C s t i I a j J                              ˆ.k K           Each dual variable i corresponds to a primal term iU . Degree of difficulty is defined as ( 1)d n m   The log-dual function   0 [ ] 1 1 ( ) :ln ( ) ln / ( ln ) ln p i i k i k i k pn i i i k k i k v V C c                         is concave on its domain of definition, and differentiable in the interior of its domain.
  • 6. Duality Theory for Composite Geometric Programming Page 5 of 27 Main Lemma of GP Lemma: If t is feasible for primal program (GP) and  is feasible for dual program (GD), then 0 ( ) ( )G Vt  . Moreover, under the same conditions 0 ( ) ( )G Vt  if, and only if, one of the following two sets of equivalent extremality conditions holds: I. ( ) 1, , (1) ˆ( ) ( ), [ ], (2) k k i k k i G k K G U i k k K              t t t II. 0( ) ( ), [0] (1) ( ), [ ], (2) i i k i U G i U i k k K         t t t in which case t is optimal for primal program (GP) and  is optimal for dual program (GD).
  • 7. Duality Theory for Composite Geometric Programming Page 6 of 27 Dual to Primal Conversion at Optimality When 0 ( ) ( )G Vt  , we get from condition II on p.5 the equations ( ), [0] ( ) , [ ], , and 0 i i i k k V i U i k k K            t  , which are log-linear:  1 ln( ( )), [0] ln , [ ], for which 0 m i ij j i j i k k V i a z c i k k K              So, knowing the optimal dual solution  an optimal primal solution can in general be easily recovered by solving this log-linear system for z and hence for t.
  • 8. Duality Theory for Composite Geometric Programming Page 7 of 27 EXAMPLE: (Maximum Likelihood Estimator of a Bernoulli Parameter) Suppose that n independent trials, each of which is a success with probability p, are performed. What is the maximum likelihood estimator (MLE) of p? The likelihood function to be maximized is 1, 1 ( , | ) (1 ) , with , 0 or 1,n n ns n s n n i if x x p p p s x x i      Example: n=3, m=2: 1 ( , ) 0 inf s.t. 1n ns s n p q f p q p q      The exponent matrix and a unique solution from extremality condition II (2) on p.5: * * 1 1 * *2 2 * 1 3 = 1 ˆ1 0 , so 0 1 n n n n n n p q n s s n s S s p p X n n n s                     
  • 9. Duality Theory for Composite Geometric Programming Page 8 of 27 Prior Extensions to GP  Signomial GP which allow some <0iC  non- convex programs: duality results are not strong  Peterson’s Generalized GP (GGP)More general separable convex programs Extensions to GP investigated in this thesis  Composite GP (CGP): itself a special case of Peterson’s GGP. It includes as special cases:  Exponential GP (EGP)  Quadratic GP (QGP)  (lpGP) We start with Exponential GP (EGP) using a motivating example…
  • 10. Duality Theory for Composite Geometric Programming Page 9 of 27 A Motivational Example (Maximum Likelihood Estimator of a Poisson Parameter) Suppose 1, , nX X are independent Poisson random variables each having mean λ. Determine the MLE of λ. The likelihood function to be maximized is 1, 1 1 ( , | ) / ( ! !), where :n ns n n n n ii f x x e x x s x          Equivalently, one can 0min ns n e      . Although this is not a posynomial, it can be solved as an EGP: 1 1 1 1 From condition (e) on p.14, we get 1 ˆ/ = , so / 1 n n n n y s y s n S n X s              This same method also works for Exponential and Normal parameters…
  • 11. Duality Theory for Composite Geometric Programming Page 10 of 27 Exponential Geometric Program (EGP) In the previous example:  a posynomial term is multiplied by an exponential factor of another posynomial term. This gives rise to an EGP problem where some posynomial term ( )iU t is multiplied by an exponential factor of another posynomial ( ): ( )i ll i E V   t t . Primal EGP problem (EGP): (cf (GP) on p.1) 0inf ( ) . . ( ) 1, ,m kR G s t G k K  0   <t t t where [ ] 1 ˆ( ) : ( ) exp ( ) , : {0} ( ) , ,and ( ) , {1, , } 1 ,| | 0, ij lj k i li k l i a b i i j l l jj J j J i n G U V k K K U C t i I V D t l L L r n i r r r r                                    t t t t t    [ ]:ljB b r m 
  • 12. Duality Theory for Composite Geometric Programming Page 11 of 27 Equivalent formulations of (EGP) In the variables A convex form of (EGP): (EGP)z 0inf ( ) . . ( ) 0,m k R g s t g k K      z z z , where   [ ] ˆ( ) :ln ( ) ln exp exp ,i k k i li k l i g G c d k K           l z t a z b z + In the variables : , : ,i lx i I l L   i l a z, b z, [ ] [ ] ( ) ( , ) with exp,k k k k k lg geo h l L      z x c ξ d where  [ ] [ ] ( , ): ln exp exp( )k k i li k l i geo x        x ξ : k kn r R R R  is called an exponential geometric function, [ ] [ ] [ ] [ ] , [ ] , :irk i i k i k l l i ii k R r r        ξ ξ ξ , and [ ]k d is similarly defined. = ln , ,j jz t j J
  • 13. Duality Theory for Composite Geometric Programming Page 12 of 27 Equivalent formulations of (EGP) (cont’d) A GGP formulation of (EGP): (EGP)x,ξ 0 0 [0] [0] ( ) [ ] [ ] inf ( , ) . . ( , ) 0, ,( ) n r R R k k k k geo s t geo k K               x, x c ξ d x c ξ d x,ξ    P where  | m n r M R R    z zP is the column space of the composite exponent matrix A B M        .
  • 14. Duality Theory for Composite Geometric Programming Page 13 of 27 Dual of EGP program (EGD) (cf (GD) on p.4) ˆ( ) ( , ) [ ] 0 1 1 sup ( , ) : . . 0, , 0, , 1, 0, (Orthogonality Conditions) 0 i l n r y i k l i R R i k l ik K i Ii l i l n r ij i lj l i l i C D y V e y s t y i I l L a y b j J y                                               , y y   [ ] 0, , (*) p.16 ˆwhere : , , : = l k i l i k l L l i i I y k K                                Note that ( ), ( )i i l ly U V  t t and degree of difficulty ( 1)d n r m    →p.4 The log-dual function (cf p.4)     ˆ [ ] ( , ) ln / ln /i i k i l l i l i k l ik K i I v y C y D y               y  is usc proper concave in( , )y  , and differentiable in the interior of its domain.
  • 15. Duality Theory for Composite Geometric Programming Page 14 of 27 Main Lemma of EGP Lemma If t is feasible for primal program (EGP) and ( , )y  is feasible for dual program (EGD), then 0 ( ) ( , )G V y t  Moreover, under the same conditions, 0 ( ) ( , )G V y t  if and only if one of the following two sets of equivalent extremality conditions hold: (cf I & II on p.5) I' Condition (e) & I ( ) 1, , ˆ( ) ( ), [ ], k k i k k i G k K y G U i k k K             t t t II' Condition (e) & II 0( ) ( ), [0] ( ), [ ], i i k i U G i y U i k k K         t t t where (e) ( ), ,l i lyV l i i I     t in which case t is optimal for primal program (EGP) and ( , )y  is optimal for dual program (EGD).  p.9 Solve the previous motivational Example.
  • 16. Duality Theory for Composite Geometric Programming Page 15 of 27 Composite Geometric Program (CGP) Primal CGP Program (CGP) (cf (EGP) on p.10) 0inf ( ) . . ( ) 1,m kR G s t G k K  0   <t t t , where [ ] ˆ( ) ( ) exp (ln ( )) ,k i l li k l i G U h V k K          t t t , :lh R R is a differentiable and strictly convex function. Note that: (QGP): if 21 2( ) ,lh l L    (lpGP): if ( ) | | / , where 1,lp l l lh p p l L    
  • 17. Duality Theory for Composite Geometric Programming Page 16 of 27 Composite Geometric Program (CGP) (cont’d) A Convex Form of (CGP): (CGP)z 0inf ( ) . . ( ) 0,m kR g s t g k K    z z z , where   [ ] ( ): ln ( ) ln exp i l k k i l li k l i g G c h d           z t a z b z A GGP Form of (CGP): (CGP)x, 0 [0] [ ] 0 ( ) inf ( , ) . . ( , ) 0, , ( )n r k k k R R f s t f k K      ξ ξ ξ x, x x x, P  with   [ ] ( , ) ( , ) : ln expk i l li k l i f geo x h              
  • 18. Duality Theory for Composite Geometric Programming Page 17 of 27 Dual program (CGD) (cf (EGD) on p.13) * ˆ( ) [ ] 0 1 1 sup ( , ): exp ( / ) . . 0, , 1, (Normality Condition) 0, (Ortho i n r y i k l l i l l i R R l ii kk K i Ii i n r ij i lj l i l C V d y h y y s t y i I a y b j J                                       y,η y  [ ] gonality Conditions) , , (**) ˆwhere : , . l i l k i i k y J l i i I y k K                            where * lh , the conjugate of lh , has domain interval lJ . (**) 0l  in (EGD) on p. 13 and that in any (CGD) (*) 0 0, ,i ly l i i I          The log-dual function (cf p.13)   * ˆ [ ] ( , ) ln / ( / )i i k i l l i l l i i k l ik K i I v y C y d y h y                 y  is usc proper concave in( , )y  , and differentiable in the interior of its domain.
  • 19. Duality Theory for Composite Geometric Programming Page 18 of 27 Main Lemma of CGP Lemma: If t is feasible for primal program (CGP) and ( , )y  is feasible for dual program (CGD), then 0 ( ) ( , )G V t y  . Moreover, under the same conditions, 0 ( ) ( , )G V t y  if and only if one of the following two sets of equivalent extremality conditions hold: (cf I’ & II’ on p.14) I'' Condition (c) & I ( ) 1, , ˆ( ) ( ), [ ], k k i k k i G k K y G U i k k K             t t t II'' Condition (c) & II 0( ) ( ), [0] ( ), [ ], i i k i U G i y U i k k K         t t t where (c) ' (ln ( )), ,ll i ly h V l i i I     t in which case t is optimal for primal program (CGP) and ( , )y  is optimal for dual program (CGD). (lpGP): (c) 1 | ln ( ) | sgn(ln ( )), ,lp l i l ly V V l i i I      t t
  • 20. Duality Theory for Composite Geometric Programming Page 19 of 27 First Duality Theorem of CGP NOTE: The statements for EGP, QGP, lpGP cases are almost the same. Theorem: Suppose that primal program (CGP) is super- consistent. Then the following three conditions are equivalent: 1) t’ is a minimal solution to (CGP). 2) There exists a vector ' p R for z′ (where z′= ln t′) such that ( ', ')z  forms a saddle point of ( , )l z  . where ( , )l z  is the Lagrangian of (CGP)z.
  • 21. Duality Theory for Composite Geometric Programming Page 20 of 27 3) There exists a vector ' p R for z′ (where z′= ln t′) such that ( ', ')z  satisfies the KKT conditions for (CGP)z: in which case the set of all such vectors 'λ is a non-empty compact convex subset of p R , and the dual program (CGD) also has a maximum solution ( ', ')y  such that 0min(CGP) ( ') ( ', ') max(CGD)G V   t y  and they satisfy the extremality conditions I''& II'' on p.18. (Perfect Duality) ' ' ' ' 0ˆ ( ) 0, ( ) 0, ( ) 0, ( ) ( , ') ( ) , 1 k k k k z k kk K a g g k K b l g where               0     z' z' z' z'
  • 22. Duality Theory for Composite Geometric Programming Page 21 of 27 Second Duality Theorem of EGP From linear programming duality theory: Of the following two linear systems, exactly one has a solution (where  : A B M n r m        ): (I) Find z with 0 0        A B z (II) Find >0 with T T A B         y y 0  We say that program (EGD) is canonical if system (II) has a solution when M is the program’s composite exponent matrix. Theorem: Suppose that primal program (EGP) is consistent. Then the minimum set of program (EGP)z is non-empty and bounded if, and only if, dual program (EGD) is canonical, in which case program (EGP) has a minimum solution t′.
  • 23. Duality Theory for Composite Geometric Programming Page 22 of 27 Second Duality Theorem of lpGP* * This theorem also applies to its special case QGP. Again, from linear programming duality theory Of the following two linear systems exactly one has a solution (where  : A B M n r m         ): (I) Find z with ,  0 Az 0 Bz 0 (II) Find with , and T T A B          y y 0 y 0   We say that program (lpGD) is canonical if system (II) has a solution when M is the program’s composite exponent matrix. Theorem 7.3.2 Suppose that primal program (lpGP) is consistent. Then the minimum set of program (lpGP)z is non-empty and bounded if and only if its dual program (lpGD) is canonical.
  • 24. Duality Theory for Composite Geometric Programming Page 23 of 27 SUMMARY OF CONTRIBUTIONS  Extension of traditional GP models to o to the more general Exponential GP models, o to the even more general Composite GP models, which include as important special cases:  (EGP)  (lpGP)  (QGP)  Showing that all of these are special cases of Peterson’s GGP models (for which he has given a Main Lemma but not a First or a Second duality theorem).
  • 25. Duality Theory for Composite Geometric Programming Page 24 of 27 CONTRIBUTIONS (cont’d)  For each extension: o multiple direct proofs of the Main Lemma, with a 2nd set of equivalent Extremality Conditions.  For the First Duality Theorem, proof that o a superconsistent primal program (CGP) has a minimal solution 't if, and only if, there exists a vector ' p R such that ( ', ')z  forms a saddle point of the Lagrangian ( , )l z  and if, and only if, it satisfies the KKT conditions for (CGP)z, o in which case the set of all such Lagrange multiplier vectors λ is a non-empty compact convex subset of p R , whereas the original theorem is a “If…then” statement and only showed the existence ofλ for the (GP) case only.
  • 26. Duality Theory for Composite Geometric Programming Page 25 of 27  For the Second Duality Theorem, proof that o for the special (and more important) cases of (QGP), (lpGP) and (EGP),  the minimum set of a consistent primal program is non-empty and bounded if and only if its dual program is canonical  the original theorem showed the existence of a primal solution for the (GP) case only, when its dual program is canonical.  Sensitivity Analysis: If * z is optimal for (CGP)z: * * * * * *0 0 0 * * * * * *0 0 ( ) ( ) ( ) , , , ( ) ( ) , , where ln i l k i l k i j l j k k ij lj g g g y c d b g g y z z b B a b                          z z z z z  Dual to primal conversion of optimal solutions when there is no duality gap: 0 ( ) ( , )G V y t  .
  • 27. Duality Theory for Composite Geometric Programming Page 26 of 27 Directions for Future Research 1.Development of proofs of a strong version of the First Duality Theorem of (CGP): If a primal program (CGP) is superconsistent and has a finite infimum, then the dual program (CGD) has a maximum solution ( ', ')y  such that inf(CGP) max(CGD) ( ', ')V  y  2.Development of computational algorithms.