tf.keras.ops.pinv

Compute the (Moore-Penrose) pseudoinverse of a matrix.

View aliases

Main aliases

tf.keras.ops.linalg.pinv

tf.keras.ops.pinv(
    x, rcond=None
)

The pseudoinverse is computed via the singular value decomposition (SVD). Singular values smaller than rcond times the largest singular value (per matrix, in the batched case) are treated as zero.

Args
x	Input tensor of shape (..., M, N).
rcond	Cutoff ratio for small singular values. Singular values less than or equal to rcond * largest_singular_value are set to zero. If None (default), the backend's own default is used.

Returns
A tensor of shape (..., N, M) containing the pseudoinverse of each matrix in x.

Example:

a = keras.ops.convert_to_tensor([[1., 2.], [3., 4.], [5., 6.]])
keras.ops.pinv(a) @ a
array([[1., 0.],
       [0., 1.]], dtype=float32)