adelie.matrix.convex_relu#

adelie.matrix.convex_relu(mat: ndarray | csc_matrix, mask: ndarray, *, gated: bool = False, copy: bool = False, n_threads: int = 1)[source]#

Creates a feature matrix for the convex relu problem.

The feature matrix for the convex gated relu problem is given by

\[\begin{align*} Y &= \begin{bmatrix} D_1 Z & \ldots & D_m Z \end{bmatrix} \end{align*}\]

where \(D_i \in \{0, 1\}^{n \times n}\) are diagonal masking matrices. The feature matrix for the convex relu problem is given by

\[\begin{align*} X &= \begin{bmatrix} Y & -Y \end{bmatrix} \end{align*}\]

Note

This matrix only works for naive method!

Parameters:

mat(n, d) Union[ndarray, csc_matrix]: The base matrix \(Z\) from which to construct the convex relu matrix.
mask(n, m) ndarray: The boolean mask matrix whose columns define the diagonal of \(D_i\). If it is not in "F"-ordering, an "F"-ordered copy is made.
gatedbool, optional: If True, the matrix will represent \(Y\) and otherwise \(X\). Default is False.
copybool, optional: If True, a copy of the inputs is stored internally. Otherwise, a reference is stored instead. Default is False.
n_threadsint, optional: Number of threads. Default is 1.

Returns:

wrap: Wrapper matrix object.