Google I/O is a wrap! Catch up on TensorFlow sessions View sessions

tf.linalg.LinearOperatorHouseholder

TensorFlow 2 version View source on GitHub

LinearOperator acting like a [batch] of Householder transformations.

Inherits From: LinearOperator

This operator acts like a [batch] of householder reflections with shape [B1,...,Bb, N, N] for some b >= 0. The first b indices index a batch member. For every batch index (i1,...,ib), A[i1,...,ib, : :] is an N x N matrix. This matrix A is not materialized, but for purposes of broadcasting this shape will be relevant.

LinearOperatorHouseholder is initialized with a (batch) vector.

A Householder reflection, defined via a vector v, which reflects points in R^n about the hyperplane orthogonal to v and through the origin.

# Create a 2 x 2 householder transform.
vec = [1 / np.sqrt(2), 1. / np.sqrt(2)]
operator = LinearOperatorHouseholder(vec)

operator.to_dense()
==> [[0.,  -1.]
     [-1., -0.]]

operator.shape
==> [2, 2]

operator.log_abs_determinant()
==> scalar Tensor

x = ... Shape [2, 4] Tensor
operator.matmul(x)
==> Shape [2, 4] Tensor

#### Shape compatibility

This operator acts on [batch] matrix with compatible shape.
`x` is a batch matrix with compatible shape for `matmul` and `solve` if

operator.shape = [B1,...,Bb] + [N, N], with b >= 0 x.shape = [C1,...,Cc] + [N, R], and [C1,...,Cc] broadcasts with [B1,...,Bb] to [D1,...,Dd]


#### Matrix property hints

This `LinearOperator` is initialized with boolean flags of the form `is_X`,
for `X = non_singular, self_adjoint, positive_definite, square`.
These have the following meaning:

* If `is_X == True`, callers should expect the operator to have the
  property `X`.  This is a promise that should be fulfilled, but is *not* a
  runtime assert.  For example, finite floating point precision may result
  in these promises being violated.
* If `is_X == False`, callers should expect the operator to not have `X`.
* If `is_X == None` (the default), callers should have no expectation either
  way.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2"><h2 class="add-link">Args</h2></th></tr>

<tr>
<td>
`reflection_axis`
</td>
<td>
Shape `[B1,...,Bb, N]` `Tensor` with `b >= 0` `N >= 0`.
The vector defining the hyperplane to reflect about.
Allowed dtypes: `float16`, `float32`, `float64`, `complex64`,
`complex128`.
</td>
</tr><tr>
<td>
`is_non_singular`
</td>
<td>
Expect that this operator is non-singular.
</td>
</tr><tr>
<td>
`is_self_adjoint`
</td>
<td>
Expect that this operator is equal to its hermitian
transpose.  This is autoset to true
</td>
</tr><tr>
<td>
`is_positive_definite`
</td>
<td>
Expect that this operator is positive definite,
meaning the quadratic form `x^H A x` has positive real part for all
nonzero `x`.  Note that we do not require the operator to be
self-adjoint to be positive-definite.  See:
<a href="https://en.wikipedia.org/wiki/Positive-definite_matrix#Extension_for_non-symmetric_matrices">https://en.wikipedia.org/wiki/Positive-definite_matrix#Extension_for_non-symmetric_matrices</a>
This is autoset to false.
</td>
</tr><tr>
<td>
`is_square`
</td>
<td>
Expect that this operator acts like square [batch] matrices.
This is autoset to true.
</td>
</tr><tr>
<td>
`name`
</td>
<td>
A name for this `LinearOperator`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2"><h2 class="add-link">Raises</h2></th></tr>

<tr>
<td>
`ValueError`
</td>
<td>
`is_self_adjoint` is not `True`, `is_positive_definite` is
not `False` or `is_square` is not `True`.
</td>
</tr>
</table>





<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2"><h2 class="add-link">Attributes</h2></th></tr>

<tr>
<td>
`H`
</td>
<td>
Returns the adjoint of the current `LinearOperator`.

Given `A` representing this `LinearOperator`, return `A*`.
Note that calling `self.adjoint()` and `self.H` are equivalent.
</td>
</tr><tr>
<td>
`batch_shape`
</td>
<td>
`TensorShape` of batch dimensions of this `LinearOperator`.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns
`TensorShape([B1,...,Bb])`, equivalent to `A.shape[:-2]`
</td>
</tr><tr>
<td>
`domain_dimension`
</td>
<td>
Dimension (in the sense of vector spaces) of the domain of this operator.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns `N`.
</td>
</tr><tr>
<td>
`dtype`
</td>
<td>
The `DType` of `Tensor`s handled by this `LinearOperator`.
</td>
</tr><tr>
<td>
`graph_parents`
</td>
<td>
List of graph dependencies of this `LinearOperator`.
</td>
</tr><tr>
<td>
`is_non_singular`
</td>
<td>

</td>
</tr><tr>
<td>
`is_positive_definite`
</td>
<td>

</td>
</tr><tr>
<td>
`is_self_adjoint`
</td>
<td>

</td>
</tr><tr>
<td>
`is_square`
</td>
<td>
Return `True/False` depending on if this operator is square.
</td>
</tr><tr>
<td>
`range_dimension`
</td>
<td>
Dimension (in the sense of vector spaces) of the range of this operator.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns `M`.
</td>
</tr><tr>
<td>
`reflection_axis`
</td>
<td>

</td>
</tr><tr>
<td>
`shape`
</td>
<td>
`TensorShape` of this `LinearOperator`.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns
`TensorShape([B1,...,Bb, M, N])`, equivalent to `A.shape`.
</td>
</tr><tr>
<td>
`tensor_rank`
</td>
<td>
Rank (in the sense of tensors) of matrix corresponding to this operator.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns `b + 2`.
</td>
</tr>
</table>



## Methods

<h3 id="add_to_tensor"><code>add_to_tensor</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L1014-L1027">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>add_to_tensor(
    x, name='add_to_tensor'
)
</code></pre>

Add matrix represented by this operator to `x`.  Equivalent to `A + x`.


<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`x`
</td>
<td>
`Tensor` with same `dtype` and shape broadcastable to `self.shape`.
</td>
</tr><tr>
<td>
`name`
</td>
<td>
A name to give this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
A `Tensor` with broadcast shape and same `dtype` as `self`.
</td>
</tr>

</table>



<h3 id="adjoint"><code>adjoint</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L870-L885">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>adjoint(
    name='adjoint'
)
</code></pre>

Returns the adjoint of the current `LinearOperator`.

Given `A` representing this `LinearOperator`, return `A*`.
Note that calling `self.adjoint()` and `self.H` are equivalent.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`LinearOperator` which represents the adjoint of this `LinearOperator`.
</td>
</tr>

</table>



<h3 id="assert_non_singular"><code>assert_non_singular</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L485-L503">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>assert_non_singular(
    name='assert_non_singular'
)
</code></pre>

Returns an `Op` that asserts this operator is non singular.

This operator is considered non-singular if

ConditionNumber < max{100, range_dimension, domain_dimension} * eps, eps := np.finfo(self.dtype.as_numpy_dtype).eps


<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A string name to prepend to created ops.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
An `Assert` `Op`, that, when run, will raise an `InvalidArgumentError` if
the operator is singular.
</td>
</tr>

</table>



<h3 id="assert_positive_definite"><code>assert_positive_definite</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L521-L536">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>assert_positive_definite(
    name='assert_positive_definite'
)
</code></pre>

Returns an `Op` that asserts this operator is positive definite.

Here, positive definite means that the quadratic form `x^H A x` has positive
real part for all nonzero `x`.  Note that we do not require the operator to
be self-adjoint to be positive definite.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name to give this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
An `Assert` `Op`, that, when run, will raise an `InvalidArgumentError` if
the operator is not positive definite.
</td>
</tr>

</table>



<h3 id="assert_self_adjoint"><code>assert_self_adjoint</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L548-L562">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>assert_self_adjoint(
    name='assert_self_adjoint'
)
</code></pre>

Returns an `Op` that asserts this operator is self-adjoint.

Here we check that this operator is *exactly* equal to its hermitian
transpose.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A string name to prepend to created ops.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
An `Assert` `Op`, that, when run, will raise an `InvalidArgumentError` if
the operator is not self-adjoint.
</td>
</tr>

</table>



<h3 id="batch_shape_tensor"><code>batch_shape_tensor</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L319-L339">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>batch_shape_tensor(
    name='batch_shape_tensor'
)
</code></pre>

Shape of batch dimensions of this operator, determined at runtime.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns a `Tensor` holding
`[B1,...,Bb]`.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`int32` `Tensor`
</td>
</tr>

</table>



<h3 id="cholesky"><code>cholesky</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L915-L938">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>cholesky(
    name='cholesky'
)
</code></pre>

Returns a Cholesky factor as a `LinearOperator`.

Given `A` representing this `LinearOperator`, if `A` is positive definite
self-adjoint, return `L`, where `A = L L^T`, i.e. the cholesky
decomposition.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`LinearOperator` which represents the lower triangular matrix
in the Cholesky decomposition.
</td>
</tr>

</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Raises</th></tr>

<tr>
<td>
`ValueError`
</td>
<td>
When the `LinearOperator` is not hinted to be positive
definite and self adjoint.
</td>
</tr>
</table>



<h3 id="determinant"><code>determinant</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L678-L695">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>determinant(
    name='det'
)
</code></pre>

Determinant for every batch member.


<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`Tensor` with shape `self.batch_shape` and same `dtype` as `self`.
</td>
</tr>

</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Raises</th></tr>

<tr>
<td>
`NotImplementedError`
</td>
<td>
If `self.is_square` is `False`.
</td>
</tr>
</table>



<h3 id="diag_part"><code>diag_part</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L965-L991">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>diag_part(
    name='diag_part'
)
</code></pre>

Efficiently get the [batch] diagonal part of this operator.

If this operator has shape `[B1,...,Bb, M, N]`, this returns a
`Tensor` `diagonal`, of shape `[B1,...,Bb, min(M, N)]`, where
`diagonal[b1,...,bb, i] = self.to_dense()[b1,...,bb, i, i]`.

my_operator = LinearOperatorDiag([1., 2.])

Efficiently get the diagonal

my_operator.diag_part() ==> [1., 2.]

Equivalent, but inefficient method

tf.linalg.diag_part(my_operator.to_dense()) ==> [1., 2.]


<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>

<tr>
<td>
`diag_part`
</td>
<td>
A `Tensor` of same `dtype` as self.
</td>
</tr>
</table>



<h3 id="domain_dimension_tensor"><code>domain_dimension_tensor</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L394-L415">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>domain_dimension_tensor(
    name='domain_dimension_tensor'
)
</code></pre>

Dimension (in the sense of vector spaces) of the domain of this operator.

Determined at runtime.

If this operator acts like the batch matrix `A` with
`A.shape = [B1,...,Bb, M, N]`, then this returns `N`.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`int32` `Tensor`
</td>
</tr>

</table>



<h3 id="inverse"><code>inverse</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L890-L913">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>inverse(
    name='inverse'
)
</code></pre>

Returns the Inverse of this `LinearOperator`.

Given `A` representing this `LinearOperator`, return a `LinearOperator`
representing `A^-1`.

<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name scope to use for ops added by this method.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`LinearOperator` representing inverse of this matrix.
</td>
</tr>

</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Raises</th></tr>

<tr>
<td>
`ValueError`
</td>
<td>
When the `LinearOperator` is not hinted to be `non_singular`.
</td>
</tr>
</table>



<h3 id="log_abs_determinant"><code>log_abs_determinant</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L707-L724">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>log_abs_determinant(
    name='log_abs_det'
)
</code></pre>

Log absolute value of determinant for every batch member.


<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Args</th></tr>

<tr>
<td>
`name`
</td>
<td>
A name for this `Op`.
</td>
</tr>
</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Returns</th></tr>
<tr class="alt">
<td colspan="2">
`Tensor` with shape `self.batch_shape` and same `dtype` as `self`.
</td>
</tr>

</table>



<!-- Tabular view -->
 <table class="responsive fixed orange">
<colgroup><col width="214px"><col></colgroup>
<tr><th colspan="2">Raises</th></tr>

<tr>
<td>
`NotImplementedError`
</td>
<td>
If `self.is_square` is `False`.
</td>
</tr>
</table>



<h3 id="matmul"><code>matmul</code></h3>

<a target="_blank" href="https://github.com/tensorflow/tensorflow/blob/v1.15.0/tensorflow/python/ops/linalg/linear_operator.py#L575-L628">View source</a>

<pre class="devsite-click-to-copy prettyprint lang-py tfo-signature-link">
<code>matmul(
    x, adjoint=False, adjoint_arg=False, name='matmul'
)
</code></pre>

Transform [batch] matrix `x` with left multiplication:  `x --> Ax`.

```python
# Make an operator acting like batch matrix A.  Assume A.shape = [..., M, N]
operator = LinearOperator(...)
operator.shape = [..., M, N]

X = ... # shape [..., N, R], batch matrix, R > 0.

Y = operator.matmul(X)
Y.shape
==> [..., M, R]

Y[..., :, r] = sum_j A[..., :, j] X[j, r]

Args
x LinearOperator or Tensor with compatible shape and same dtype as self. See class docstring for definition of compatibility.
adjoint Python bool. If True, left multiply by t