You are on page 1of 18

Session 3

Semidefinite Programming (SDP)


&
Second-Order Cone Programming (SOCP)
Relation with other optimization
problems
Semidefinite Programming (SDP)
• Definition: Semidefinite programming (SDP) is a class of
convex optimization where optimization function is linear
and the constraints are either linear equalities/inequalities or
generalized inequalities with respect to semidefinite cone.
• SDP is solvable via interior point methods
• Semidefinite programming is an extension of linear
programming where the componentwise inequalities
between vectors are replaced by matrix inequalities, or,
equivalently the first orthant is replaced by the cone of
positive semi-definite matrices.
 If All Eigen values of M ≥ 0
 If All Principal Minors ≥ 0
 If All Pivot values ≥ 0

06/20/2022 4
 If All Eigen values of M ≤ 0
 If All Principal Minors ≤ 0
 If All Pivot values ≤ 0

06/20/2022 5
Tests1 of Semidefinite matrices

06/20/2022 6
The determinant of a principal submatrix is called the principal minor of A.

06/20/2022 7
06/20/2022 8
06/20/2022 9
SDP Standard form
• A semidefinite optimization problem has the form

• Here F0, F1,…, Fm are symmetric matrices in ℝm× m and 

•  means that F(x) is positive semidefinite.

• Such a constraint is nonlinear, but convex; therefore semidefinite programs are convex
optimization problems. 
Difference between LP and SDP
Primal and Dual form of SDP
Example
Second-Order Cone Programming (SOCPs)
• Why SOCP??
• SOCPs can be solved by interior point
methods and in general, can be solved more
efficiently than semidefinite programming (SDP)
problems. 
• It's called the second order cone because it's
defined by a quadratic equation rather than a
linear one.
• This naming is consistent with a quadratic
polynomials being of degree two while a linear
polynomial is of degree one.
06/20/2022 14
Second-Order Cone Programming (SOCPs)

• The second-order cone in   is the set of vectors   with 


• The picture shows part of the cone, and a slice (at  y=1). For some strange reason,
this set is called an ‘‘ice-cream cone’’.
• Second-order cone programming (SOCP) is a generalization of linear and quadratic
programming that allows for affine combination of variables to be constrained
inside second-order cones. The SOCP model includes as special cases problems with
convex quadratic objective and constraints.
• SOCP models are particularly useful in geometry problems, as well as in linear
programs where the data is imprecisely known.

06/20/2022 15
General Form of SOCPs
SOCP-Example : Robust Linear Program

06/20/2022 17
References
• https://
cvxopt.org/userguide/coneprog.html#seco
nd-order-cone-programming
• https://
cvxopt.org/userguide/coneprog.html#semi
definite-programming
• https://
docs.mosek.com/modeling-cookbook/linea
r.html
• https://nptel.ac.in/courses/111/104/111104
068
06/20/2022 18

You might also like