- chainerx.linalg.svd(a, full_matrices=True, compute_uv=True)¶
Singular Value Decomposition.
Factorizes the matrix
ainto two unitary matrices
Vt, and a 1-D array
sof singular values such that
a == U * S * Vt, where
Sis a suitably shaped matrix of zeros with main diagonal
*represents a dot product.
A tuple of
(U, s, Vt)such that
a = U * diag(s) * Vt. When
compute_uvis False only singular values
- Return type
float16is not supported yet.)
The SVD is commonly written as a = U * diag(s) * V^T. The
Vtreturned by this function is V^T.
During backpropagation, this function requires
Vtcomputed, therefore differentiation does not work for
Backpropagation is not implemented for