gemmr.generative_model._generate_random_dominant_subspace_rotations¶
-
gemmr.generative_model.
_generate_random_dominant_subspace_rotations
(U, V, m, qx, qy, rng, uvrots)¶ Generates random weight vectors in the dominant subspaces (of dimension qx and qy).
Separately for X and Y, random rotation matrices are obtained via QR decomposition of a matrix with random entries drawn from a standard normal distribution and choosing the Q factor; the basis vectors U (for X) and V (for Y are then multiplied by these rotation matrices and the first m columns are selected and returned.
Parameters: - U (np.ndarray(px, px)) – basis vectors for X
- V (np.ndarray(py, py)) – basis vectors for Y
- m (int) – number of between-set association modes
- qx (int) – the dimensionality of the “dominant” subspace for X
- qy (int) – the dimensionality of the “dominant” subspace for Y
- rng (random number generator instance) –
- uvrots (IGNORED) – present for signature compatibility with
_generate_dominant_subspace_rotations_from_opti()
Returns: - U_dominant (np.ndarray (px, m)) – weight vectors for X living in “dominant” subspace
- V_dominant (np.ndarray (py, m)) – weight vectors for Y living in “dominant subspace