KonaMatrix (base class)

class kona.linalg.matrices.common.KonaMatrix(primal=None, state=None, transposed=False)

Bases: object

An abstract matrix class connected to Kona memory. This class is used to define a variety of jacobian matrices and other composite objects containing matrix-related methods used in optimization tasks.

Parameters:

• primal (DesignVector) –
• state (StateVector) –
• transposed (boolean, optional) –

Variables:

• _design (PrimalVector) – Primal vector point for linearization.
• _state (StateVector) – State vector point for linearization
• _transposed (boolean) – Flag to determine if the matrix is transposed

T

Returns the transposed version of the matrix.

Returns:KonaMatrix-like Return type:Transposed version of the matrix.

linearize(primal, state)

Store the vector points around which a non-linear matrix should be linearized.

Parameters:

• primal (DesignVector or CompositePrimalVector) –
• state (StateVector) –

product(in_vec, out_vec)

Performs a matrix-vector product at the internally stored linearization.

Parameters:

• in_vec (KonaVector) –
• out_vec (KonaVector) –

Returns:

out_vec

Return type:

KonaVector