picos.expressions.cone_psd

Implements the positive semidefinite cone.

Classes

class picos.expressions.cone_psd.PositiveSemidefiniteCone(dim=None)[source]

Bases: Cone

The positive semidefinite cone.

Unlike other cones which are defined only on \mathbb{K}^n, this cone accepts both symmetric and hermitian matrices as well as their special vectorization as members.

Example

>>> from picos import Constant, PositiveSemidefiniteCone
>>> R = Constant("R", range(16), (4, 4))
>>> S = R + R.T
>>> S.shape
(4, 4)
>>> S.svec.shape
(10, 1)
>>> S.svec << PositiveSemidefiniteCone()  # Constrain the matrix via svec().
<4×4 LMI Constraint: R + Rᵀ ≽ 0>
>>> C = S << PositiveSemidefiniteCone(); C  # Constrain the matrix directly.
<4×4 LMI Constraint: R + Rᵀ ≽ 0>
>>> C.conic_membership_form[0]      # The conic form still refers to svec().
<10×1 Real Constant: svec(R + Rᵀ)>
>>> C.conic_membership_form[1]
<10-dim. Positive Semidefinite Cone: {svec(A) : xᵀ·A·x ≥ 0 ∀ x}>
__init__(dim=None)[source]

Construct a PositiveSemidefiniteCone.

If a fixed dimensionality is given, this must be the dimensiona of the special vectorization. For a n \times n matrix, this is \frac{n(n + 1)}{2}.

property dual_cone

Implement cone.Cone.dual_cone.