Link Search Menu Expand Document
enuSpace-tensorflow Manual

SparseFillEmptyRowsGrad

tensorflow C++ API

tensorflow::ops::SparseFillEmptyRowsGrad

The gradient of SparseFillEmptyRows.

Summary

Takes vectors reverse_index_map, shaped[N], and grad_values, shaped[N_full], whereN_full >= Nand copies data into eitherd_valuesord_default_value. Hered_valuesis shaped[N]andd_default_valueis a scalar.

d_values[j] = grad_values[reverse_index_map[j]] d_default_value = sum_{k : 0 .. N_full - 1} ( grad_values[k] * 1{k not in reverse_index_map})

Arguments:

  • scope: A Scope object
  • reverse_index_map: 1-D. The reverse index map from SparseFillEmptyRows.
  • grad_values: 1-D. The gradients from backprop.

Returns:

  • Output d_values: 1-D. The backprop into values.
  • Output d_default_value: 0-D. The backprop into default_value.

SparseFillEmptyRowsGrad block

Source link : https://github.com/EXPNUNI/enuSpaceTensorflow/blob/master/enuSpaceTensorflow/tf_sparse.cpp

Argument:

  • Scope scope : A Scope object (A scope is generated automatically each page. A scope is not connected.)
  • Input reverse_index_map: connect Input node.
  • Input grad_values: connect Input node.

Return:

  • Output d_values: Output object of SparseFillEmptyRowsGrad class object.
  • Output d_default_value: Output object of SparseFillEmptyRowsGrad class object.

Result:

  • std::vector(Tensor) result_d_values : Returned object of executed result by calling session.
  • std::vector(Tensor) result_d_default_value : Returned object of executed result by calling session.

Using Method