Method of TTTensor.
Convert a Tensor-Train representation into explicit structure and parameter matrices.
Source: src/tensor/@TTTensor/computeSPhi.m
[S, Phi] = computeSPhi(obj)computeSPhi reconstructs the explicit multilinear
structure matrix S and parameter matrix Phi
from a Tensor-Train (TT) representation.
The method enumerates all multilinear monomials represented by the TTTensor through
sequential contractions of the TT cores. To improve computational
efficiency, the TT network is split into left and right halves, which
are contracted independently and combined to generate the complete set
of monomials.
The resulting matrices are equivalent to the explicit MTI parameter representation and can be used for model conversion, analysis and compatibility with CP-based algorithms.
| Argument | Description |
|---|---|
obj |
TTTensor
object. |
tol |
Numerical threshold for removing negligible coefficients (default
1e-10). |
varOrder |
Variable ordering used to construct the structure matrix (default
1:nVar). |
bitDirection |
Bit ordering used for canonical monomial sorting, "lsb"
or "msb" (default "lsb"). |
| Output | Description |
|---|---|
S |
Sparse logical structure matrix. |
Phi |
Sparse parameter matrix. |
monomialBits |
Binary representation of each monomial. |
monomialMask |
Integer mask defining the canonical monomial order. |
TTTensor ยท computeIncidenceMatrix
MyToolbox Documentation | Generated automatically by CI/CD pipeline