Adds {batch_}matrix_solve and {batch_}triangular_matrix_solve ops.

Change: 112289410

1 files changed, 106 insertions, 0 deletions

diff --git a/tensorflow/core/kernels/matrix_solve_op.cc b/tensorflow/core/kernels/matrix_solve_op.cc
new file mode 100644
index 0000000000..992af9eb66
--- /dev/null
+++ b/tensorflow/core/kernels/matrix_solve_op.cc
@@ -0,0 +1,106 @@
+/* Copyright 2015 Google Inc. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+// See docs in ../ops/linalg_ops.cc.
+// TODO(rmlarsen): Add optional hint params so the caller can promise that the
+// matrices are invertible, symmetric (maybe detect automatically?), and
+// positive definite, which will allow us to call progressively faster solvers
+// internally.
+#include <cmath>
+
+#include "third_party/eigen3/Eigen/LU"
+#include "tensorflow/core/framework/kernel_def_builder.h"
+#include "tensorflow/core/framework/op_kernel.h"
+#include "tensorflow/core/kernels/binary_linalg_ops_common.h"
+#include "tensorflow/core/lib/core/errors.h"
+#include "tensorflow/core/platform/logging.h"
+#include "tensorflow/core/platform/port.h"
+#include "tensorflow/core/public/tensor_shape.h"
+
+namespace tensorflow {
+
+template <class Scalar, bool SupportsBatchOperationT>
+class MatrixSolveOp
+    : public BinaryLinearAlgebraOp<Scalar, SupportsBatchOperationT> {
+ public:
+  explicit MatrixSolveOp(OpKernelConstruction* context)
+      : BinaryLinearAlgebraOp<Scalar, SupportsBatchOperationT>(context) {}
+  ~MatrixSolveOp() override {}
+
+  TensorShape GetOutputMatrixShape(
+      const TensorShape& input_matrix_shape,
+      const TensorShape& rhs_matrix_shape) override {
+    CHECK_EQ(input_matrix_shape.dims(), rhs_matrix_shape.dims());
+    TensorShape output_matrix_shape = input_matrix_shape;
+    output_matrix_shape.set_dim(
+        output_matrix_shape.dims() - 1,
+        rhs_matrix_shape.dim_size(output_matrix_shape.dims() - 1));
+    return output_matrix_shape;
+  }
+
+  int64 GetCostPerUnit(const TensorShape& input_matrix_shape,
+                       const TensorShape& rhs_matrix_shape) override {
+    const int64 rows = input_matrix_shape.dim_size(0);
+    const int64 rhss = rhs_matrix_shape.dim_size(1);
+    if (rows > (1LL << 20)) {
+      // A big number to cap the cost in case overflow.
+      return kint32max;
+    } else {
+      return rows * rows * (rows + rhss);
+    }
+  }
+
+  using typename BinaryLinearAlgebraOp<Scalar, SupportsBatchOperationT>::Matrix;
+  using typename BinaryLinearAlgebraOp<Scalar,
+                                       SupportsBatchOperationT>::MatrixMap;
+  using typename BinaryLinearAlgebraOp<Scalar,
+                                       SupportsBatchOperationT>::ConstMatrixMap;
+
+  void ComputeMatrix(OpKernelContext* context, const ConstMatrixMap& matrix,
+                     const ConstMatrixMap& rhs, MatrixMap* output) override {
+    OP_REQUIRES(context, matrix.rows() == matrix.cols(),
+                errors::InvalidArgument("Input matrix must be square."));
+    OP_REQUIRES(
+        context, matrix.rows() == rhs.rows(),
+        errors::InvalidArgument("Input matrix and rhs are incompatible."));
+    if (matrix.rows() == 0) {
+      // To be consistent with the MatrixInverse op, we define the solution for
+      // an empty set of equation is the empty matrix.
+      return;
+    }
+    Eigen::PartialPivLU<Matrix> lu_decomposition(matrix);
+    // While PartialPivLU cannot give strong guarantees on invertability,
+    // we can at least guard against exact zero pivots. This can occur as
+    // a result of basic user mistakes such providing integer valued
+    // matrices that are exacly singular, or due to underflow if this
+    // code is run with denormals being flushed to zero.
+    // TODO(rmlarsen): Add check based on condition number estimation.
+    const Scalar min_abs_pivot =
+        lu_decomposition.matrixLU().diagonal().cwiseAbs().minCoeff();
+    OP_REQUIRES(context, min_abs_pivot > Scalar(0),
+                errors::InvalidArgument("Input matrix is not invertible."));
+    *output = lu_decomposition.solve(rhs);
+  }
+};
+
+REGISTER_BINARY_LINALG_OP("MatrixSolve", (MatrixSolveOp<float, false>), float);
+REGISTER_BINARY_LINALG_OP("MatrixSolve", (MatrixSolveOp<double, false>),
+                          double);
+REGISTER_BINARY_LINALG_OP("BatchMatrixSolve", (MatrixSolveOp<float, true>),
+                          float);
+REGISTER_BINARY_LINALG_OP("BatchMatrixSolve", (MatrixSolveOp<double, true>),
+                          double);
+
+}  // namespace tensorflow