LRN¶
Description¶
This module does not implement any operations, but setves as the basis for local response normalization modules.
There are two types of local normalization layers:
- inter-channel normalization;
- intra-channel normalization.
The only difference between them is the area of normalization. In the intra-channel LRN, a 2D neighborhood is determined within one channel, whereas in the inter-channel LRN, the spatial neighborhood is limited to one element, while normalization occurs along a certain neighborhood of the channels.
This normalization technique in general practically does not affect the final result, therefore it is considered obsolete.
Initializing¶
def __init__(self, N=5, alpha=1e-4, beta=0.75, K=2.0, name=None):
Parameters
| Parameter | Allowed types | Description | Default |
|---|---|---|---|
| N | int | Number of elements to normalize | 5 |
| alpha | float | Bias | 1e-4 |
| beta | float | Scale parameter | 0.75 |
| K | float | Incremental coefficient | 2.0 |
| name | str | Layer name | None |
Explanations
See descendant classes.