fromSparseComponents

Method of CPNTensor.

Static factory that builds a CPNTensor directly from its six split components.

Source: src/tensor/@CPNTensor/CPNTensor.m

Syntax

obj = CPNTensor.fromSparseComponents( ...
        structureMatrixTrue = ST, ...
        structureMatrixFalse = SF, ...
        structureMatrixContinuous = SC, ...
        parameterMatrixOne = PO, ...
        parameterMatrixMinusOne = PM, ...
        parameterMatrixContinuous = PC)

Description

fromSparseComponents assembles a CPNTensor from the six pre-split sparse components, without deriving them from a single signed matrix. It is the name–value equivalent of the six-argument constructor. Each component defaults to [], so any omitted part is empty.

The three structure components are stored as sparse matrices (True and False as logical masks, Continuous as numeric weights), and likewise the three parameter components (One and MinusOne logical, Continuous numeric). Their meanings are:

Component	Factor / coefficient
`structureMatrixTrue`	monomial factor $v_l$
`structureMatrixFalse`	complemented factor $1-v_l$
`structureMatrixContinuous`	general factor $1-\lvert s\rvert + s\,v_l$
`parameterMatrixOne`	coefficient $+1$
`parameterMatrixMinusOne`	coefficient $-1$
`parameterMatrixContinuous`	any other coefficient value

After assembly, mustHaveDisjointStructure verifies that the three structure components have non-overlapping support (each entry sourced from at most one of True / False / Continuous); it errors with the offending (row, column) positions otherwise. The tensor rank R is then set from the structure matrix.

Use this factory when the components are already separated — for example, when a model needs a structureMatrixFalse (complemented factors $1-v_l$ ), which the signed-matrix routes never populate.

Input arguments

All are name–value arguments, each defaulting to [].

Argument	Description
`structureMatrixTrue`	Logical mask of factors $v_l$ , signals by monomials.
`structureMatrixFalse`	Logical mask of factors $1-v_l$ , signals by monomials.
`structureMatrixContinuous`	Numeric weights $s$ for general factors, signals by monomials.
`parameterMatrixOne`	Logical mask of coefficient $+1$ , equations by monomials.
`parameterMatrixMinusOne`	Logical mask of coefficient $-1$ , equations by monomials.
`parameterMatrixContinuous`	Numeric remaining coefficients, equations by monomials.

Output arguments

Output	Description
`obj`	`CPNTensor` with the given components and rank `R`.

fromSparseComponents

Syntax

Description

Input arguments

Output arguments

See also