pyhgf.utils.weight_initialisation.sparse_init#

pyhgf.utils.weight_initialisation.sparse_init(n_parents, n_children, sparsity=0.9, std=0.01, seed=None)[source]#

Sparse initialisation.

Most weights are set to zero; only a fraction 1 - sparsity of entries are drawn from \(\\mathcal{N}(0, \\text{std}^2)\).

Parameters:

  • n_parents (int) – Number of parent (input) nodes — fan-in.

  • n_children (int) – Number of child (output) nodes — fan-out.

  • sparsity (float) – Fraction of weights set to zero (default 0.9 → 90 % zeros).

  • std (float) – Standard deviation of the non-zero entries (default 0.01).

  • seed (int | None) – Optional random seed for reproducibility.

Returns:

Weight vector of length n_parents * n_children.

Return type:

numpy.ndarray