MVN
Input
x:
- Required: yes
- Type: float16, float32
- Description: input tensor
- Restriction: Must be of type float16.
Attribute
- normalize_variance:
- Required: no
- Type: bool
- Description: If False, normalizes mean only (default = True).
- Restriction: either True or False
- across_channels:
- Required: no
- Type: bool
- Description: If true, CHW is regarded as a vector (default = false).
- Restriction: either True or False
- eps:
- Required: no
- Type: float
- Description: epsilon for not dividing by zero (default = 1e – 9)
- Restriction: The value is user-defined.
Output
y:
- Required: yes
- Type: float16, float32
- Description: y = (x – mean(x))/std(x)
- Restriction: Must be of type float16.
Parent topic: Supported Caffe Operators