flopscope.

flopscope.numpy.matvec

fnp.matvec(a, b, **kwargs)[flopscope source][numpy source]

Matrix-vector dot product of two arrays.

Adapted from NumPy docs np.matvec

Areacore
Typecounted
NumPy Refnp.matvec
Cost
numel(output)\text{numel}(\text{output})
Flopscope Context

Matrix-vector product. Cost = output_size * contracted_axis.

Given a matrix (or stack of matrices) A\mathbf{A} in x1 and a vector (or stack of vectors) v\mathbf{v} in x2, the matrix-vector product is defined as:

A⋅b=∑j=0n−1Aijvj\mathbf{A} \cdot \mathbf{b} = \sum_{j=0}^{n-1} A_{ij} v_j

where the sum is over the last dimensions in x1 and x2 (unless axes is specified). (For a matrix-vector product with the vector conjugated, use flops.vecmat(x2, x1.mT).)

Added in version 2.2.0.

Parameters

x1, x2:array_like

Input arrays, scalars not allowed.

out:ndarray, optional

A location into which the result is stored. If provided, it must have the broadcasted shape of x1 and x2 with the summation axis removed. If not provided or None, a freshly-allocated array is used.

**kwargs

For other keyword-only arguments, see the ufunc docs.

Returns

y:ndarray

The matrix-vector product of the inputs.

Raises

:ValueError

If the last dimensions of x1 and x2 are not the same size.

If a scalar value is passed in.

See also

Examples

Rotate a set of vectors from Y to X along Z.

>>> a = flops.array([[0., 1., 0.],
... [-1., 0., 0.],
... [0., 0., 1.]])
>>> v = flops.array([[1., 0., 0.],
... [0., 1., 0.],
... [0., 0., 1.],
... [0., 6., 8.]])
>>> flops.matvec(a, v)
array([[ 0., -1.,  0.],
       [ 1.,  0.,  0.],
       [ 0.,  0.,  1.],
       [ 6.,  0.,  8.]])