Add broadcasting to all LinearOperators.
authorA. Unique TensorFlower <gardener@tensorflow.org>
Fri, 13 Apr 2018 22:23:08 +0000 (15:23 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Fri, 13 Apr 2018 22:25:38 +0000 (15:25 -0700)
commita77dcb5e56dbbbcc3383cb0b39cd79dd88135635
tree1716ba71ef9e71d643c15f763e681b1d94413ed3
parent1298c3240aa9f36b79ea7f0e772edfff87381771
Add broadcasting to all LinearOperators.

This will broadcast in cases where batch shapes are not equal (but tries to determine statically if this is the case). The broadcasting is not as efficient as doing the broadcast in C++, but makes for the API to at least be completely broadcastable.

PiperOrigin-RevId: 192832919
tensorflow/contrib/linalg/BUILD
tensorflow/contrib/linalg/python/kernel_tests/linear_operator_block_diag_test.py
tensorflow/python/ops/linalg/linear_operator.py
tensorflow/python/ops/linalg/linear_operator_full_matrix.py
tensorflow/python/ops/linalg/linear_operator_low_rank_update.py
tensorflow/python/ops/linalg/linear_operator_lower_triangular.py
tensorflow/python/ops/linalg/linear_operator_test_util.py