Computes softmax cross entropy cost and gradients to backpropagate.
Unlike `SoftmaxCrossEntropyWithLogits`, this operation does not accept a matrix of label probabilities, but rather a single label per row of features. This label is considered to have probability 1.0 for the given row.
Inputs are the logits, not probabilities.
Constants
String | OP_NAME | The name of this op, as known by TensorFlow core engine |
Public Methods
Output<T> |
backprop()
backpropagated gradients (batch_size x num_classes matrix).
|
static <T extends TNumber> SparseSoftmaxCrossEntropyWithLogits<T> | |
Output<T> |
loss()
Per example loss (batch_size vector).
|
Inherited Methods
Constants
public static final String OP_NAME
The name of this op, as known by TensorFlow core engine
Public Methods
public static SparseSoftmaxCrossEntropyWithLogits<T> create (Scope scope, Operand<T> features, Operand<? extends TNumber> labels)
Factory method to create a class wrapping a new SparseSoftmaxCrossEntropyWithLogits operation.
Parameters
scope | current scope |
---|---|
features | batch_size x num_classes matrix |
labels | batch_size vector with values in [0, num_classes). This is the label for the given minibatch entry. |
Returns
- a new instance of SparseSoftmaxCrossEntropyWithLogits