View source on GitHub
|
Normalizes along dimension axis using an L2 norm. (deprecated arguments)
tf.compat.v1.linalg.l2_normalize(
x, axis=None, epsilon=1e-12, name=None, dim=None
)
For a 1-D tensor with axis = 0, computes
output = x / sqrt(max(sum(x**2), epsilon))
For x with more dimensions, independently normalizes each 1-D slice along
dimension axis.
Args | |
|---|---|
x
|
A Tensor.
|
axis
|
Dimension along which to normalize. A scalar or a vector of integers. |
epsilon
|
A lower bound value for the norm. Will use sqrt(epsilon) as the
divisor if norm < sqrt(epsilon).
|
name
|
A name for this operation (optional). |
dim
|
Deprecated alias for axis. |
Returns | |
|---|---|
A Tensor with the same shape as x.
|