Expressions

CVXPY represents mathematical objects as expression trees. An expression tree is a collection of mathematical expressions linked together by one or more atoms. Expression trees are encoded as instances of the Expression class, and each Leaf in a tree is a Variable, Parameter, or Constant.

Expression

class cvxpy.expressions.expression.Expression[source]

Bases: Canonical

A mathematical expression in a convex optimization problem.

Overloads many operators to allow for convenient creation of compound expressions (e.g., the sum of two expressions) and constraints.

property T

The transpose of the expression.

Type:

Expression

__add__(other: Expression) Expression[source]

Expression : Sum two expressions.

__div__(other: Expression) Expression[source]

Expression : One expression divided by another.

__eq__(other: Expression)[source]

Equality : Creates a constraint self == other.

__ge__(other: Expression)[source]

NonPos : Creates an inequality constraint.

__le__(other: Expression)[source]

Inequality : Creates an inequality constraint self <= other.

__lshift__(other: Expression) PSD[source]

PSD : Creates a negative semidefinite inequality.

__matmul__(other: Expression) Expression[source]

Expression : Matrix multiplication of two expressions.

__mul__(other: Expression) Expression[source]

Expression : The product of two expressions.

__pow__(power: float) Expression[source]

Raise expression to a power.

Parameters:

power (float) – The power to which to raise the expression.

Returns:

The expression raised to power.

Return type:

Expression

__radd__(other: Expression) Expression[source]

Expression : Sum two expressions.

__rdiv__(other: Expression) Expression[source]

Expression : Called for Number / Expression.

__rlshift__(other: Expression) PSD[source]

PSD : Creates a negative semidefinite inequality.

__rmatmul__(other: Expression) Expression[source]

Expression : Called for matrix @ Expression.

__rmul__(other: Expression) Expression[source]

Expression : Called for Number * Expression.

__rrshift__(other: Expression) PSD[source]

PSD : Creates a positive semidefinite inequality.

__rshift__(other: Expression) PSD[source]

PSD : Creates a positive semidefinite inequality.

__rsub__(other: Expression) Expression[source]

Expression : The difference of two expressions.

__rtruediv__(other: Expression) Expression[source]

Expression : Called for Number / Expression.

__sub__(other: Expression) Expression[source]

Expression : The difference of two expressions.

__truediv__(other: Expression) Expression[source]

Expression : One expression divided by another.

property curvature: str

The curvature of the expression.

Type:

str

abstract property domain

The constraints describing the closure of the region where the expression is finite.

Type:

list

abstract property grad

Gives the (sub/super)gradient of the expression w.r.t. each variable.

Matrix expressions are vectorized, so the gradient is a matrix.

Returns:

A map of variable to SciPy CSC sparse matrix; None if a variable value is missing.

Return type:

dict

is_affine() bool[source]

Is the expression affine?

abstract is_concave() bool[source]

Is the expression concave?

is_constant() bool[source]

Is the expression constant?

abstract is_convex() bool[source]

Is the expression convex?

is_dcp(dpp: bool = False) bool[source]

Checks whether the Expression is DCP.

Parameters:

dpp (bool, optional) – If True, enforce the disciplined parametrized programming (DPP) ruleset; only relevant when the problem involves Parameters.

Returns:

True if the Expression is DCP, False otherwise.

Return type:

bool

is_dgp(dpp: bool = False) bool[source]

Checks whether the Expression is log-log DCP.

Returns:

True if the Expression is log-log DCP, False otherwise.

Return type:

bool

abstract is_dpp(context: str = 'dcp') bool[source]

The expression is a disciplined parameterized expression.

is_dqcp() bool[source]

Checks whether the Expression is DQCP.

Returns:

True if the Expression is DQCP, False otherwise.

Return type:

bool

is_log_log_affine() bool[source]

Is the expression affine?

abstract is_log_log_concave() bool[source]

Is the expression log-log concave?

abstract is_log_log_convex() bool[source]

Is the expression log-log convex?

abstract is_nonneg() bool[source]

Is the expression positive?

abstract is_nonpos() bool[source]

Is the expression negative?

is_zero() bool[source]

Is the expression all zero?

abstract name() str[source]

str : The string representation of the expression.

property ndim: int

The number of dimensions in the expression’s shape.

Type:

int

abstract property shape: Tuple[int, ...]

The expression dimensions.

Type:

tuple

property sign: str

The sign of the expression.

Type:

str

property size: int

The number of entries in the expression.

Type:

int

abstract property value

The numeric value of the expression.

Type:

NumPy.ndarray or None

Leaf

class cvxpy.expressions.leaf.Leaf(shape: Union[int, Tuple[int, ...]], value=None, nonneg: bool = False, nonpos: bool = False, complex: bool = False, imag: bool = False, symmetric: bool = False, diag: bool = False, PSD: bool = False, NSD: bool = False, hermitian: bool = False, boolean: bool = False, integer: bool = False, sparsity=None, pos: bool = False, neg: bool = False)[source]

Bases: Expression

A leaf node of an expression tree; i.e., a Variable, Constant, or Parameter.

A leaf may carry attributes that constrain the set values permissible for it. Leafs can have no more than one attribute, with the exception that a leaf may be both nonpos and nonneg or both boolean in some indices and integer in others.

An error is raised if a leaf is assigned a value that contradicts one or more of its attributes. See the project method for a convenient way to project a value onto a leaf’s domain.

Parameters:
  • shape (tuple or int) – The leaf dimensions. Either an integer n for a 1D shape, or a tuple where the semantics are the same as NumPy ndarray shapes. Shapes cannot be more than 2D.

  • value (numeric type) – A value to assign to the leaf.

  • nonneg (bool) – Is the variable constrained to be nonnegative?

  • nonpos (bool) – Is the variable constrained to be nonpositive?

  • complex (bool) – Is the variable complex valued?

  • symmetric (bool) – Is the variable symmetric?

  • diag (bool) – Is the variable diagonal?

  • PSD (bool) – Is the variable constrained to be positive semidefinite?

  • NSD (bool) – Is the variable constrained to be negative semidefinite?

  • Hermitian (bool) – Is the variable Hermitian?

  • boolean (bool or list of tuple) – Is the variable boolean? True, which constrains the entire Variable to be boolean, False, or a list of indices which should be constrained as boolean, where each index is a tuple of length exactly equal to the length of shape.

  • integer (bool or list of tuple) – Is the variable integer? The semantics are the same as the boolean argument.

  • sparsity (list of tuplewith) – Fixed sparsity pattern for the variable.

  • pos (bool) – Is the variable positive?

  • neg (bool) – Is the variable negative?

property T

The transpose of the expression.

Type:

Expression

property ndim: int

The number of dimensions in the expression’s shape.

Type:

int

project(val)[source]

Project value onto the attribute set of the leaf.

A sensible idiom is leaf.value = leaf.project(val).

Parameters:

val (numeric type) – The value assigned.

Returns:

The value rounded to the attribute type.

Return type:

numeric type

project_and_assign(val) None[source]

Project and assign a value to the variable.

property shape: Tuple[int, ...]

The dimensions of the expression.

Type:

tuple

property size: int

The number of entries in the expression.

Type:

int

property value

The numeric value of the parameter.

Type:

NumPy.ndarray or None

Variable

class cvxpy.expressions.variable.Variable(shape=(), name: Optional[str] = None, var_id: Optional[int] = None, **kwargs: Any)[source]

Bases: Leaf

The optimization variables in a problem.

property T

The transpose of the expression.

Type:

Expression

name() str[source]

str : The name of the variable.

property ndim: int

The number of dimensions in the expression’s shape.

Type:

int

project(val)

Project value onto the attribute set of the leaf.

A sensible idiom is leaf.value = leaf.project(val).

Parameters:

val (numeric type) – The value assigned.

Returns:

The value rounded to the attribute type.

Return type:

numeric type

project_and_assign(val) None

Project and assign a value to the variable.

property shape: Tuple[int, ...]

The dimensions of the expression.

Type:

tuple

property size: int

The number of entries in the expression.

Type:

int

property value

The numeric value of the parameter.

Type:

NumPy.ndarray or None

Parameter

class cvxpy.expressions.constants.parameter.Parameter(shape: Tuple[int, ...] = (), name: Optional[str] = None, value=None, id=None, **kwargs)[source]

Bases: Leaf

Parameters in optimization problems.

Parameters are constant expressions whose value may be specified after problem creation. The only way to modify a problem after its creation is through parameters. For example, you might choose to declare the hyper-parameters of a machine learning model to be Parameter objects; more generally, Parameters are useful for computing trade-off curves.

property T

The transpose of the expression.

Type:

Expression

property ndim: int

The number of dimensions in the expression’s shape.

Type:

int

project(val)

Project value onto the attribute set of the leaf.

A sensible idiom is leaf.value = leaf.project(val).

Parameters:

val (numeric type) – The value assigned.

Returns:

The value rounded to the attribute type.

Return type:

numeric type

project_and_assign(val) None

Project and assign a value to the variable.

property shape: Tuple[int, ...]

The dimensions of the expression.

Type:

tuple

property size: int

The number of entries in the expression.

Type:

int

property value

The numeric value of the parameter.

Type:

NumPy.ndarray or None

Constant

class cvxpy.expressions.constants.Constant(value)[source]

Bases: Leaf

A constant value.

Raw numerical constants (Python primite types, NumPy ndarrays, and NumPy matrices) are implicitly cast to constants via Expression operator overloading. For example, if x is an expression and c is a raw constant, then x + c creates an expression by casting c to a Constant.

property T

The transpose of the expression.

Type:

Expression

property ndim: int

The number of dimensions in the expression’s shape.

Type:

int

property shape: Tuple[int, ...]

Returns the (row, col) dimensions of the expression.

property size: int

The number of entries in the expression.

Type:

int

property value

The numeric value of the constant.

Type:

NumPy.ndarray or None