SlideShare a Scribd company logo
1 of 38
Download to read offline
Convex	
  optimization	
  problems
guodong
Outline
• Optimization	
  problem
• Convex	
  optimization	
  problem
• Linear	
  Programming	
  (LP)
• Quadratic	
  Programming	
  (QP)
• Quadratic	
  Constraint	
  Quadratic	
  Programming	
  (QCQP)
• Second-­‐order	
  cone	
  programming	
  (SOCP)
• Geometric	
  Programming
• Generalized	
  inequality	
  constraints
– Semi-­‐definite	
  programming
• Vector	
  optimization
Convex	
  optimization	
  solvers	
  are	
  not	
  covered
• Convex	
  optimization	
  problem	
  involved	
  today
– LP:	
  perfectly	
  solved	
  since	
  1948;	
  
• Simplex	
  algorithm(1947);	
  dual(1947);	
  dual	
  simplex	
  (1954);	
  Episode	
  algorithm	
  (1975,	
  
polynomial	
  time);	
  
– QP:	
  solving	
  difficulty	
  is	
  close	
  with	
  LP	
  (for	
  convex	
  QP)
• Interior	
  point	
  method	
  ,	
  active	
  set	
  method,	
  conjugate	
  gradient	
  method
– SOCP:	
  ‘has	
  very	
  efficient	
  method	
  to	
  solve	
  SOCP’
• Tools/Library
– CVX:	
  convex	
  optimization	
  library	
  on	
  Matlab
– CVXOPT:	
  free	
  library	
  on	
  Python
– Joptimizer:	
  on	
  Java
Optimization	
  problem	
  
• Standard	
  form
• Domain	
  and	
  implicit	
  constraints
• Feasible
– Variable	
  x	
  is	
  feasible:	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  and	
  x	
  satisfice	
  all	
  constraints
– Problem	
  is	
  feasible:	
  at	
  least	
  exists	
  one	
  feasible	
  variable
x ∈D
Optimal	
  value	
  and	
  optimal	
  solution
• Optimal	
  value
• Optimal	
  solution
• Local	
  optimal	
  point
4	
  Results	
  for	
  one	
  optimal	
  problem
• Not	
  feasible
• Unbounded	
  below
• Has	
  optimal	
  value,	
  but	
  no	
  optimal	
  point
• Has	
  optimal	
  value,	
  and	
  also	
  optimal	
  point
f1(x) = ex
Convex	
  optimization
• Stanford	
  form
• Feasible	
  set	
  of	
  a	
  convex	
  optimization	
  problem	
  is	
  convex	
  set
– Domain	
  of	
  convex	
  function	
  is	
  convex	
  set
– Sublevel	
  set	
  of	
  convex	
  function	
  is	
  convex	
  set	
  
– Intersection	
  of	
  convex	
  sets	
  are	
  convex	
  set
Abstract	
  convex	
  optimization	
  problem
• Convex	
  optimization:	
  Minimize	
  a	
  convex	
  function	
  in	
  convex	
  set
• Convert	
  to	
  a	
  equivalent	
  convex	
  problem	
  may	
  help	
  solve
Local	
  optima	
  is	
  global	
  optima
– Consider	
  
– We	
  have:	
  
– We	
  can	
  easily	
  choose	
  a	
  small	
  	
  	
  	
  ,	
  which	
  contradicts	
  our	
  assumption	
  that	
  
x	
  is	
  local	
  optimal
• E.g.	
  
θ
Optimality	
  criterion	
  for	
  differentiable	
  f0
• X is optimal if and only if it’s feasible and
• It’s easy to prove x is optimal if	
  above	
  conditionholds
• Prove	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  if	
  x is optimal solution
– X	
  is	
  interior	
  of	
  feasible	
  set	
  (constraints	
  do	
  not	
  work	
  )
– X	
  is	
  in	
  boundary	
  of	
  feasible	
  set	
  (constraints	
  probably	
  works)
∇f0 (x) = 0
∇f0 (x) ≠ 0
∇f0 (x)T
(y − x) ≥ 0 for all feasible y
we have: f0 (y) ≥ f0 (x)+ ∇f0 (x)T
(y − x)
∇f0 (x)T
(y − x) ≥ 0, ∀y ∈X
Optimality	
  criterion	
  for	
  differentiable	
  f0
• If	
  x	
  is	
  optimal	
  and	
  in	
  the	
  boundary	
  (	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  ),	
  then
∇f0 (x)T
(y − x) ≥ 0, ∀y ∈X
∇f0 (x) ≠ 0
Assume for some y, ∇f0 (x)T
(y − x) < 0
z(t) = ty + (1-t)x, t ∈[0,1]
d
dt
f0 (z(t))|t=0 = ∇f0 (x)T
(y − x) < 0
f0 (z(t)) < f0 (x)
Equivalent	
  convex	
  problem:	
  examples
• Two	
  problems	
  are	
  equivalent	
  if	
  the	
  solution	
  of	
  one	
  is	
  readily	
  
obtained	
  from	
  the	
  the	
  other	
  one
– Some	
  transformations	
  preserve	
  convexity,	
  which	
  is	
  nice
• Example:	
  eliminating	
  equality	
  constraints
Equivalent	
  convex	
  problem:	
  examples
• Introducing	
  equality	
  constraints
• Introducing	
  slack	
  variables	
  for	
  linear	
  inequalities
Equivalent	
  convex	
  problem:	
  examples
Linear	
  Programming	
  (LP)
• Definition
– Convex	
  problem	
  with	
  affine	
  objective	
  and	
  constraint	
  functions
• Geometric	
  interpretation
– Feasible	
  set	
  is	
  a	
  polyhedron
– Level	
  set	
  is	
  hyper-­‐plane
– Level	
  set	
  is	
  orthogonal	
  with	
  c
– Optimal	
  solution	
  usually	
  appear	
  in	
  vertex
G ∈Rm×n
,A ∈Rp×n
LP	
  example	
  (1)
•
LP	
  example	
  (2)
• Chebyshev	
  center	
  of	
  a	
  polyhedron
B = {xc + u | || u ||2 ≤ r}
ai
T
x ≤ bi for all x ∈B if and only if
sup{ai
T
(xc + u)| || u ||2 ≤ r}=sup{ai
T
xc + || ai ||2|| u ||2 )| || u ||2 ≤ r}=ai
T
xc + r || ai ||2 ≤ bi
Linear	
  fractional	
  program
• Standard	
  form
• Transform	
  to	
  LP
y =
x
eT
x + f
, z=
1
eT
x + f
if y, z are solution of the new problem, then x = y / z is the solution of original problem
and also have same objective function
if x is solution of the original problem, then y and z are the solution of new problem
and have same objective function cT
y + dz = f0 (x)
Quadratic	
  program	
  (QP)
• Standard	
  form
– Minimize	
  a	
  convex quadratic	
  function	
  in	
  a	
  polyhedron
• Geometric	
  interpretation
– Feasible	
  set	
  is	
  a	
  polyhedron
– Level	
  set	
  is	
  ellipsoid
– Optimal	
  solution	
  usually	
  appear	
  in	
  boundary	
  
P ∈S+
n
QP	
  examples
• Least	
  squares	
  with	
  bound	
  constraints
• Distance	
  between	
  two	
  polyhedron
minimize || Ax -b ||2
2
= xT
AT
Ax - 2bT
Ax + bT
b
subject to li ≤ xi ≤ ui , i = 1,...,n
AT
A ∈S+
n
Minimize || x1 - x2 ||2
2
, subject to A1x1 ≤ b1, A2x2 ≤ b2
Quadratically	
  constrained	
  quadratic	
  program	
  (QCQP)
• Standard	
  form
• LP ⊆ QP ⊆ QCQP
Second-­‐order	
  cone	
  program	
  
i ||⋅||2 is Euclidean norm || y ||2 = y1
2
+...+ yn
2
i constraints are nonlinear, nondifferentiable, convex
Second order cone: {(z,t)| zT
z ≤ t2
,t ≥ 0}
SOCP	
  Example:	
  Parameter	
  uncertainty	
  in	
  LP
•
• Deterministic	
  model
– contains	
  must	
  hold	
  for	
  all	
  
ai ∈εi = {ai + Piu | || u ||≤1}, Pi ∈Rn×n
ai ∈εi
sup{ai
T
x | ai ∈εi } ≤ bi
sup{ai
T
x | ai ∈εi } = ai
T
x + sup{uT
Pi
T
x | ||u ||≤1} = ai
T
x+ || Pi
T
x ||2
(when u =
Pi
T
x
|| Pi
T
x ||2
)
So ai
T
x+ || Pi
T
x ||2 ≤ b, which is SOCP
SOCP	
  Example:	
  Parameter	
  uncertainty	
  in	
  L
• Stochastic	
  model
ai is random variable; constraints must hold with probability η
bi − ai
T
x
|| ⋅i
1/2
∑ x ||
≥ Φ−1
(η)
Geometric	
  Programming
• Monomial	
  function
• Posynomial	
  function:	
  sum	
  of	
  monomial	
  function
• Operations
– Monomial	
  function:	
  multiple/divide	
  number;	
  multiple/divide	
  another	
  
monomial	
  function
– Posynomial	
  function:	
  multiple/divide	
  number;	
  add/multiple/divide	
  
monomial	
  function;	
  add	
  another	
  posynomial	
  function
Geometric	
  Programming
• Geometric	
  Programming
• Example
minimize x / y
subject to 2 ≤ x ≤ 3
x2
+ 3y / z ≤ y
x / y = z2
Geometric	
  Programming	
  in	
  convex	
  form
• Idea:
• Details
Generalized	
  inequality	
  constraint
• Convex	
  problem	
  with	
  generalized	
  inequality	
  constraints
• Same	
  properties	
  are	
  standard	
  convex	
  problem
– Convex	
  feasible	
  set;	
  local	
  optimal	
  is	
  global;	
  optimality	
  criterion	
  for	
  
differentiable	
  f0
when Ki = R+
n
, it's a normal convex problem
Cone	
  form	
  problem	
  
• Special	
  case	
  with	
  affine	
  objective	
  and	
  constraints
• SOCP	
  is	
  a	
  special	
  case	
  of	
  cone	
  form	
  problem
minimize cT
x
subject to -(Ai x + bi ,ci
T
x + di ) =Ki
0, i = 1,...,m
Fx = g,
in which, Ki = {(y,t) ∈Rni +1
| || y ||2 ≤ t} is a second order cone
Review:	
  Cone
• Cone
• Convex	
  cone
• Positive	
  semi-­‐definite	
  cone
• Proper	
  cone
• Dual	
  cone
if for any x ∈C and θ ≥ 0, we have θx ∈C, then we say set C is cone
if for any x1,x2 ∈C and θ1,θ2 ≥ 0, always have θ1x1 +θ2x2 ∈C
set of all positive semi - definite matrix : S+
n
= {X ∈Sn
| X = 0}
a cone which is convex, closed, solid, and pointed
dual cone of a cone K : K*
= {y | yT
x ≥ 0 for all x ∈K}
Review:	
  Generalized	
  inequality	
  
• Define	
  partial	
  order	
  using	
  proper	
  cone	
  K
• Popular	
  proper	
  cones	
  and	
  corresponding	
  generalized	
  
inequalities
• Dual	
  property
x =K y <=> y - x ∈K
x K y <=> y - x ∈int K
K = R+ , then =K is ≤ in R
K = R+
n
, then x =K y equals xi ≤ yi ,i = 1,...,n
K = S+
n
, then X =K Y equals Y - X is PSDM
x =K y iff for any λ =K* 0 and λ ≠ 0, we have λT
x < λT
y
Semi-­‐definite	
  programming
• Definition
LP	
  and	
  SOCP	
  as	
  SDP
Vector	
  optimization
• General	
  vector	
  optimization	
  problem
• Convex	
  vector	
  optimization	
  problem
Optimal	
  and	
  Pareto	
  optimal	
  point
• Set	
  of	
  achievable	
  objective	
  values
Scalarization
x =K y iff for any λ =K 0 and λ ≠ 0, we have λT
x < λT
y
Multi	
  criterion	
  optimization	
  
• Vector	
  optimization	
  problem	
  with K = R+
q
Example:	
  Regularized	
  least-­‐squares

More Related Content

What's hot

The extended kalman filter
The extended kalman filterThe extended kalman filter
The extended kalman filterMudit Parnami
 
Calculus of variation problems
Calculus of variation   problemsCalculus of variation   problems
Calculus of variation problemsSolo Hermelin
 
Multi Objective Optimization
Multi Objective OptimizationMulti Objective Optimization
Multi Objective OptimizationNawroz University
 
Eigenvalues and Eigenvectors
Eigenvalues and EigenvectorsEigenvalues and Eigenvectors
Eigenvalues and EigenvectorsVinod Srivastava
 
Diagonalization of Matrices
Diagonalization of MatricesDiagonalization of Matrices
Diagonalization of MatricesAmenahGondal1
 
Fuzzy relations
Fuzzy relationsFuzzy relations
Fuzzy relationsnaugariya
 
Numerical analysis m1 l3slides
Numerical analysis  m1 l3slidesNumerical analysis  m1 l3slides
Numerical analysis m1 l3slidesSHAMJITH KM
 
Eigen values and eigen vectors engineering
Eigen values and eigen vectors engineeringEigen values and eigen vectors engineering
Eigen values and eigen vectors engineeringshubham211
 
Eigen value , eigen vectors, caley hamilton theorem
Eigen value , eigen vectors, caley hamilton theoremEigen value , eigen vectors, caley hamilton theorem
Eigen value , eigen vectors, caley hamilton theoremgidc engineering college
 
Reinforcement learning
Reinforcement learning Reinforcement learning
Reinforcement learning Chandra Meena
 
Classical relations and fuzzy relations
Classical relations and fuzzy relationsClassical relations and fuzzy relations
Classical relations and fuzzy relationsBaran Kaynak
 
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE ShivangiSingh241
 
Alpha-beta pruning (Artificial Intelligence)
Alpha-beta pruning (Artificial Intelligence)Alpha-beta pruning (Artificial Intelligence)
Alpha-beta pruning (Artificial Intelligence)Falak Chaudry
 

What's hot (20)

Special functions
Special functionsSpecial functions
Special functions
 
The extended kalman filter
The extended kalman filterThe extended kalman filter
The extended kalman filter
 
FUZZY COMPLEMENT
FUZZY COMPLEMENTFUZZY COMPLEMENT
FUZZY COMPLEMENT
 
Edge detection
Edge detectionEdge detection
Edge detection
 
Introduction to optimization Problems
Introduction to optimization ProblemsIntroduction to optimization Problems
Introduction to optimization Problems
 
Calculus of variation problems
Calculus of variation   problemsCalculus of variation   problems
Calculus of variation problems
 
Multi Objective Optimization
Multi Objective OptimizationMulti Objective Optimization
Multi Objective Optimization
 
Eigenvalues and Eigenvectors
Eigenvalues and EigenvectorsEigenvalues and Eigenvectors
Eigenvalues and Eigenvectors
 
Diagonalization of Matrices
Diagonalization of MatricesDiagonalization of Matrices
Diagonalization of Matrices
 
Fuzzy relations
Fuzzy relationsFuzzy relations
Fuzzy relations
 
Numerical analysis m1 l3slides
Numerical analysis  m1 l3slidesNumerical analysis  m1 l3slides
Numerical analysis m1 l3slides
 
Recursion DM
Recursion DMRecursion DM
Recursion DM
 
Eigen values and eigen vectors engineering
Eigen values and eigen vectors engineeringEigen values and eigen vectors engineering
Eigen values and eigen vectors engineering
 
Genetic Algorithms
Genetic AlgorithmsGenetic Algorithms
Genetic Algorithms
 
Eigen value , eigen vectors, caley hamilton theorem
Eigen value , eigen vectors, caley hamilton theoremEigen value , eigen vectors, caley hamilton theorem
Eigen value , eigen vectors, caley hamilton theorem
 
Reinforcement learning
Reinforcement learning Reinforcement learning
Reinforcement learning
 
Classical relations and fuzzy relations
Classical relations and fuzzy relationsClassical relations and fuzzy relations
Classical relations and fuzzy relations
 
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE
DIGITAL COMMUNICATION: ENCODING AND DECODING OF CYCLIC CODE
 
Alpha-beta pruning (Artificial Intelligence)
Alpha-beta pruning (Artificial Intelligence)Alpha-beta pruning (Artificial Intelligence)
Alpha-beta pruning (Artificial Intelligence)
 
Image segmentation
Image segmentation Image segmentation
Image segmentation
 

Similar to Convex optimization methods

super vector machines algorithms using deep
super vector machines algorithms using deepsuper vector machines algorithms using deep
super vector machines algorithms using deepKNaveenKumarECE
 
Convex Optimization Modelling with CVXOPT
Convex Optimization Modelling with CVXOPTConvex Optimization Modelling with CVXOPT
Convex Optimization Modelling with CVXOPTandrewmart11
 
Lp and ip programming cp 9
Lp and ip programming cp 9Lp and ip programming cp 9
Lp and ip programming cp 9M S Prasad
 
Support Vector Machines is the the the the the the the the the
Support Vector Machines is the the the the the the the the theSupport Vector Machines is the the the the the the the the the
Support Vector Machines is the the the the the the the the thesanjaibalajeessn
 
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...Yandex
 
Optimization introduction
Optimization introductionOptimization introduction
Optimization introductionhelalmohammad2
 
Elliptical curve cryptography
Elliptical curve cryptographyElliptical curve cryptography
Elliptical curve cryptographyBarani Tharan
 
lecture01_lecture01_lecture0001_ceva.pdf
lecture01_lecture01_lecture0001_ceva.pdflecture01_lecture01_lecture0001_ceva.pdf
lecture01_lecture01_lecture0001_ceva.pdfAnaNeacsu5
 
Nonconvex Compressed Sensing with the Sum-of-Squares Method
Nonconvex Compressed Sensing with the Sum-of-Squares MethodNonconvex Compressed Sensing with the Sum-of-Squares Method
Nonconvex Compressed Sensing with the Sum-of-Squares MethodTasuku Soma
 

Similar to Convex optimization methods (20)

super vector machines algorithms using deep
super vector machines algorithms using deepsuper vector machines algorithms using deep
super vector machines algorithms using deep
 
Convex Optimization Modelling with CVXOPT
Convex Optimization Modelling with CVXOPTConvex Optimization Modelling with CVXOPT
Convex Optimization Modelling with CVXOPT
 
bv_cvxslides (1).pdf
bv_cvxslides (1).pdfbv_cvxslides (1).pdf
bv_cvxslides (1).pdf
 
Lp and ip programming cp 9
Lp and ip programming cp 9Lp and ip programming cp 9
Lp and ip programming cp 9
 
Support Vector Machines is the the the the the the the the the
Support Vector Machines is the the the the the the the the theSupport Vector Machines is the the the the the the the the the
Support Vector Machines is the the the the the the the the the
 
Boyd 4.6, 4.7
Boyd 4.6, 4.7Boyd 4.6, 4.7
Boyd 4.6, 4.7
 
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...
Subgradient Methods for Huge-Scale Optimization Problems - Юрий Нестеров, Cat...
 
QMC: Operator Splitting Workshop, A Splitting Method for Nonsmooth Nonconvex ...
QMC: Operator Splitting Workshop, A Splitting Method for Nonsmooth Nonconvex ...QMC: Operator Splitting Workshop, A Splitting Method for Nonsmooth Nonconvex ...
QMC: Operator Splitting Workshop, A Splitting Method for Nonsmooth Nonconvex ...
 
Optimization introduction
Optimization introductionOptimization introduction
Optimization introduction
 
Elliptical curve cryptography
Elliptical curve cryptographyElliptical curve cryptography
Elliptical curve cryptography
 
lecture01_lecture01_lecture0001_ceva.pdf
lecture01_lecture01_lecture0001_ceva.pdflecture01_lecture01_lecture0001_ceva.pdf
lecture01_lecture01_lecture0001_ceva.pdf
 
Nonconvex Compressed Sensing with the Sum-of-Squares Method
Nonconvex Compressed Sensing with the Sum-of-Squares MethodNonconvex Compressed Sensing with the Sum-of-Squares Method
Nonconvex Compressed Sensing with the Sum-of-Squares Method
 
DAA.pdf
DAA.pdfDAA.pdf
DAA.pdf
 
DAA.pdf
DAA.pdfDAA.pdf
DAA.pdf
 
Optimization tutorial
Optimization tutorialOptimization tutorial
Optimization tutorial
 
L20 Simplex Method
L20 Simplex MethodL20 Simplex Method
L20 Simplex Method
 
ECC_basics.ppt
ECC_basics.pptECC_basics.ppt
ECC_basics.ppt
 
ECC_basics.ppt
ECC_basics.pptECC_basics.ppt
ECC_basics.ppt
 
Introduction to Sparse Methods
Introduction to Sparse Methods Introduction to Sparse Methods
Introduction to Sparse Methods
 
Optim_methods.pdf
Optim_methods.pdfOptim_methods.pdf
Optim_methods.pdf
 

More from Dong Guo

AlphaGo zero
AlphaGo zeroAlphaGo zero
AlphaGo zeroDong Guo
 
DQN (Deep Q-Network)
DQN (Deep Q-Network)DQN (Deep Q-Network)
DQN (Deep Q-Network)Dong Guo
 
机器学习概述
机器学习概述机器学习概述
机器学习概述Dong Guo
 
Expectation propagation
Expectation propagationExpectation propagation
Expectation propagationDong Guo
 
Additive model and boosting tree
Additive model and boosting treeAdditive model and boosting tree
Additive model and boosting treeDong Guo
 
Feature selection
Feature selectionFeature selection
Feature selectionDong Guo
 
Logistic Regression
Logistic RegressionLogistic Regression
Logistic RegressionDong Guo
 
Machine learning Introduction
Machine learning IntroductionMachine learning Introduction
Machine learning IntroductionDong Guo
 

More from Dong Guo (8)

AlphaGo zero
AlphaGo zeroAlphaGo zero
AlphaGo zero
 
DQN (Deep Q-Network)
DQN (Deep Q-Network)DQN (Deep Q-Network)
DQN (Deep Q-Network)
 
机器学习概述
机器学习概述机器学习概述
机器学习概述
 
Expectation propagation
Expectation propagationExpectation propagation
Expectation propagation
 
Additive model and boosting tree
Additive model and boosting treeAdditive model and boosting tree
Additive model and boosting tree
 
Feature selection
Feature selectionFeature selection
Feature selection
 
Logistic Regression
Logistic RegressionLogistic Regression
Logistic Regression
 
Machine learning Introduction
Machine learning IntroductionMachine learning Introduction
Machine learning Introduction
 

Recently uploaded

SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024Lorenzo Miniero
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Commit University
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr LapshynFwdays
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii SoldatenkoFwdays
 
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024BookNet Canada
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptxLBM Solutions
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDGMarianaLemus7
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationSlibray Presentation
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Enterprise Knowledge
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsMark Billinghurst
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr BaganFwdays
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitecturePixlogix Infotech
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Scott Keck-Warren
 
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Wonjun Hwang
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Neo4j
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfAddepto
 

Recently uploaded (20)

SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024
 
Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!Nell’iperspazio con Rocket: il Framework Web di Rust!
Nell’iperspazio con Rocket: il Framework Web di Rust!
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko
 
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
Transcript: New from BookNet Canada for 2024: BNC BiblioShare - Tech Forum 2024
 
Key Features Of Token Development (1).pptx
Key  Features Of Token  Development (1).pptxKey  Features Of Token  Development (1).pptx
Key Features Of Token Development (1).pptx
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
APIForce Zurich 5 April Automation LPDG
APIForce Zurich 5 April  Automation LPDGAPIForce Zurich 5 April  Automation LPDG
APIForce Zurich 5 April Automation LPDG
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck Presentation
 
Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024Designing IA for AI - Information Architecture Conference 2024
Designing IA for AI - Information Architecture Conference 2024
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR Systems
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping Elbows
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC Architecture
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024
 
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
Bun (KitWorks Team Study 노별마루 발표 2024.4.22)
 
Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024Build your next Gen AI Breakthrough - April 2024
Build your next Gen AI Breakthrough - April 2024
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
Gen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdfGen AI in Business - Global Trends Report 2024.pdf
Gen AI in Business - Global Trends Report 2024.pdf
 

Convex optimization methods

  • 2. Outline • Optimization  problem • Convex  optimization  problem • Linear  Programming  (LP) • Quadratic  Programming  (QP) • Quadratic  Constraint  Quadratic  Programming  (QCQP) • Second-­‐order  cone  programming  (SOCP) • Geometric  Programming • Generalized  inequality  constraints – Semi-­‐definite  programming • Vector  optimization
  • 3. Convex  optimization  solvers  are  not  covered • Convex  optimization  problem  involved  today – LP:  perfectly  solved  since  1948;   • Simplex  algorithm(1947);  dual(1947);  dual  simplex  (1954);  Episode  algorithm  (1975,   polynomial  time);   – QP:  solving  difficulty  is  close  with  LP  (for  convex  QP) • Interior  point  method  ,  active  set  method,  conjugate  gradient  method – SOCP:  ‘has  very  efficient  method  to  solve  SOCP’ • Tools/Library – CVX:  convex  optimization  library  on  Matlab – CVXOPT:  free  library  on  Python – Joptimizer:  on  Java
  • 4. Optimization  problem   • Standard  form • Domain  and  implicit  constraints • Feasible – Variable  x  is  feasible:                          and  x  satisfice  all  constraints – Problem  is  feasible:  at  least  exists  one  feasible  variable x ∈D
  • 5. Optimal  value  and  optimal  solution • Optimal  value • Optimal  solution • Local  optimal  point
  • 6. 4  Results  for  one  optimal  problem • Not  feasible • Unbounded  below • Has  optimal  value,  but  no  optimal  point • Has  optimal  value,  and  also  optimal  point f1(x) = ex
  • 7. Convex  optimization • Stanford  form • Feasible  set  of  a  convex  optimization  problem  is  convex  set – Domain  of  convex  function  is  convex  set – Sublevel  set  of  convex  function  is  convex  set   – Intersection  of  convex  sets  are  convex  set
  • 8. Abstract  convex  optimization  problem • Convex  optimization:  Minimize  a  convex  function  in  convex  set • Convert  to  a  equivalent  convex  problem  may  help  solve
  • 9. Local  optima  is  global  optima – Consider   – We  have:   – We  can  easily  choose  a  small        ,  which  contradicts  our  assumption  that   x  is  local  optimal • E.g.   θ
  • 10. Optimality  criterion  for  differentiable  f0 • X is optimal if and only if it’s feasible and • It’s easy to prove x is optimal if  above  conditionholds • Prove                                                            if  x is optimal solution – X  is  interior  of  feasible  set  (constraints  do  not  work  ) – X  is  in  boundary  of  feasible  set  (constraints  probably  works) ∇f0 (x) = 0 ∇f0 (x) ≠ 0 ∇f0 (x)T (y − x) ≥ 0 for all feasible y we have: f0 (y) ≥ f0 (x)+ ∇f0 (x)T (y − x) ∇f0 (x)T (y − x) ≥ 0, ∀y ∈X
  • 11. Optimality  criterion  for  differentiable  f0 • If  x  is  optimal  and  in  the  boundary  (                              ),  then ∇f0 (x)T (y − x) ≥ 0, ∀y ∈X ∇f0 (x) ≠ 0 Assume for some y, ∇f0 (x)T (y − x) < 0 z(t) = ty + (1-t)x, t ∈[0,1] d dt f0 (z(t))|t=0 = ∇f0 (x)T (y − x) < 0 f0 (z(t)) < f0 (x)
  • 12. Equivalent  convex  problem:  examples • Two  problems  are  equivalent  if  the  solution  of  one  is  readily   obtained  from  the  the  other  one – Some  transformations  preserve  convexity,  which  is  nice • Example:  eliminating  equality  constraints
  • 13. Equivalent  convex  problem:  examples • Introducing  equality  constraints • Introducing  slack  variables  for  linear  inequalities
  • 15. Linear  Programming  (LP) • Definition – Convex  problem  with  affine  objective  and  constraint  functions • Geometric  interpretation – Feasible  set  is  a  polyhedron – Level  set  is  hyper-­‐plane – Level  set  is  orthogonal  with  c – Optimal  solution  usually  appear  in  vertex G ∈Rm×n ,A ∈Rp×n
  • 17. LP  example  (2) • Chebyshev  center  of  a  polyhedron B = {xc + u | || u ||2 ≤ r} ai T x ≤ bi for all x ∈B if and only if sup{ai T (xc + u)| || u ||2 ≤ r}=sup{ai T xc + || ai ||2|| u ||2 )| || u ||2 ≤ r}=ai T xc + r || ai ||2 ≤ bi
  • 18. Linear  fractional  program • Standard  form • Transform  to  LP y = x eT x + f , z= 1 eT x + f if y, z are solution of the new problem, then x = y / z is the solution of original problem and also have same objective function if x is solution of the original problem, then y and z are the solution of new problem and have same objective function cT y + dz = f0 (x)
  • 19. Quadratic  program  (QP) • Standard  form – Minimize  a  convex quadratic  function  in  a  polyhedron • Geometric  interpretation – Feasible  set  is  a  polyhedron – Level  set  is  ellipsoid – Optimal  solution  usually  appear  in  boundary   P ∈S+ n
  • 20. QP  examples • Least  squares  with  bound  constraints • Distance  between  two  polyhedron minimize || Ax -b ||2 2 = xT AT Ax - 2bT Ax + bT b subject to li ≤ xi ≤ ui , i = 1,...,n AT A ∈S+ n Minimize || x1 - x2 ||2 2 , subject to A1x1 ≤ b1, A2x2 ≤ b2
  • 21. Quadratically  constrained  quadratic  program  (QCQP) • Standard  form • LP ⊆ QP ⊆ QCQP
  • 22. Second-­‐order  cone  program   i ||⋅||2 is Euclidean norm || y ||2 = y1 2 +...+ yn 2 i constraints are nonlinear, nondifferentiable, convex Second order cone: {(z,t)| zT z ≤ t2 ,t ≥ 0}
  • 23. SOCP  Example:  Parameter  uncertainty  in  LP • • Deterministic  model – contains  must  hold  for  all   ai ∈εi = {ai + Piu | || u ||≤1}, Pi ∈Rn×n ai ∈εi sup{ai T x | ai ∈εi } ≤ bi sup{ai T x | ai ∈εi } = ai T x + sup{uT Pi T x | ||u ||≤1} = ai T x+ || Pi T x ||2 (when u = Pi T x || Pi T x ||2 ) So ai T x+ || Pi T x ||2 ≤ b, which is SOCP
  • 24. SOCP  Example:  Parameter  uncertainty  in  L • Stochastic  model ai is random variable; constraints must hold with probability η bi − ai T x || ⋅i 1/2 ∑ x || ≥ Φ−1 (η)
  • 25. Geometric  Programming • Monomial  function • Posynomial  function:  sum  of  monomial  function • Operations – Monomial  function:  multiple/divide  number;  multiple/divide  another   monomial  function – Posynomial  function:  multiple/divide  number;  add/multiple/divide   monomial  function;  add  another  posynomial  function
  • 26. Geometric  Programming • Geometric  Programming • Example minimize x / y subject to 2 ≤ x ≤ 3 x2 + 3y / z ≤ y x / y = z2
  • 27. Geometric  Programming  in  convex  form • Idea: • Details
  • 28. Generalized  inequality  constraint • Convex  problem  with  generalized  inequality  constraints • Same  properties  are  standard  convex  problem – Convex  feasible  set;  local  optimal  is  global;  optimality  criterion  for   differentiable  f0 when Ki = R+ n , it's a normal convex problem
  • 29. Cone  form  problem   • Special  case  with  affine  objective  and  constraints • SOCP  is  a  special  case  of  cone  form  problem minimize cT x subject to -(Ai x + bi ,ci T x + di ) =Ki 0, i = 1,...,m Fx = g, in which, Ki = {(y,t) ∈Rni +1 | || y ||2 ≤ t} is a second order cone
  • 30. Review:  Cone • Cone • Convex  cone • Positive  semi-­‐definite  cone • Proper  cone • Dual  cone if for any x ∈C and θ ≥ 0, we have θx ∈C, then we say set C is cone if for any x1,x2 ∈C and θ1,θ2 ≥ 0, always have θ1x1 +θ2x2 ∈C set of all positive semi - definite matrix : S+ n = {X ∈Sn | X = 0} a cone which is convex, closed, solid, and pointed dual cone of a cone K : K* = {y | yT x ≥ 0 for all x ∈K}
  • 31. Review:  Generalized  inequality   • Define  partial  order  using  proper  cone  K • Popular  proper  cones  and  corresponding  generalized   inequalities • Dual  property x =K y <=> y - x ∈K x K y <=> y - x ∈int K K = R+ , then =K is ≤ in R K = R+ n , then x =K y equals xi ≤ yi ,i = 1,...,n K = S+ n , then X =K Y equals Y - X is PSDM x =K y iff for any λ =K* 0 and λ ≠ 0, we have λT x < λT y
  • 33. LP  and  SOCP  as  SDP
  • 34. Vector  optimization • General  vector  optimization  problem • Convex  vector  optimization  problem
  • 35. Optimal  and  Pareto  optimal  point • Set  of  achievable  objective  values
  • 36. Scalarization x =K y iff for any λ =K 0 and λ ≠ 0, we have λT x < λT y
  • 37. Multi  criterion  optimization   • Vector  optimization  problem  with K = R+ q