Computes the exponential linear function.
tf.nn.elu(
features: Annotated[Any, TV_Elu_T], name=None
) -> Annotated[Any, TV_Elu_T]
The ELU function is defined as:
- \( e ^ x - 1 \) if \( x < 0 \)
- \( x \) if \( x >= 0 \)
Examples:
tf.nn.elu(1.0)<tf.Tensor: shape=(), dtype=float32, numpy=1.0>tf.nn.elu(0.0)<tf.Tensor: shape=(), dtype=float32, numpy=0.0>tf.nn.elu(-1000.0)<tf.Tensor: shape=(), dtype=float32, numpy=-1.0>
See Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
Args | |
|---|---|
features
|
A Tensor. Must be one of the following types: half, bfloat16, float32, float64.
|
name
|
A name for the operation (optional). |
Returns | |
|---|---|
A Tensor. Has the same type as features.
|