Developer Guide for Intel® Data Analytics Acceleration Library 2018 Update 3
The backward local contrast normalization layer accepts the input described below. Pass the Input ID as a parameter to the methods that provide input for your algorithm. For more details, see Algorithms.
Result ID |
Result |
|
---|---|---|
inputGradient |
Pointer to tensor G of size n1 x n2 x n3 x n4 that stores the input gradient computed on the preceding layer. This input can be an object of any class derived from Tensor. |
|
inputFromForward |
Collection of input data needed for the backward local contrast normalization layer. This collection can contain objects of any class derived from Tensor. |
|
Element ID |
Element |
|
auxInvMax |
Pointer to the tensor:
This tensor stores the inverted max values. This input can be an object of any class derived from Tensor. |
|
auxCenteredData |
Pointer to tensor x(5)nqij of size n1 x n2 x n3 x n4 that stores values as shown above. This input can be an object of any class derived from Tensor. |
|
auxSigma |
Pointer to tensor x(9)nij of size n1 x n3 x n4 if sumDimension is not NULL, or tensor x(9)nqij of size n1 x n2 x n3 x n4 otherwise, that stores values as shown above. This input can be an object of any class derived from Tensor. |
|
auxC |
Pointer to tensor x(12)n of size n1 if sumDimension is not NULL, or tensor x(12)nq of size n1 x n3 otherwise, that stores values as shown above. This input can be an object of any class derived from Tensor. |
For common parameters of neural network layers, see Common Parameters.
In addition to the common parameters, the backward local contrast normalization layer has the following parameters:
Parameter |
Default Value |
Description |
|
---|---|---|---|
algorithmFPType |
float |
The floating-point type that the algorithm uses for intermediate computations. Can be float or double. |
|
method |
defaultDense |
Default computation method used by the algorithm, the only method supported by the layer. |
|
kernel |
HomogenTensor<float> of size 5 x 5 with values 0.04 |
Tensor with sizes m1 x m2 of the two-dimensional kernel. Only kernels with odd dimensions are currently supported. |
|
indices |
indices(2,3) |
Data structure representing dimensions k1 and k2 for kernels. |
|
sumDimension |
HomogenNumericTable<float> of size 1 x 1 with value 1 |
Numeric table of size 1 x 1 that stores dimension f. If it is NULL, there is no summation over this dimension. |
|
sigmaDegenerateCasesThreshold |
1e-04 |
The threshold to avoid degenerate cases when calculating σ-1. |
The backward local contrast normalization layer calculates the result described below. Pass the Result ID as a parameter to the methods that access the results of your algorithm. For more details, see Algorithms.
Input ID |
Input |
|
---|---|---|
gradient |
Pointer to tensor Z of size n1 x n2 x n3 x n4 that stores the result of the backward local contrast normalization layer. This input can be an object of any class derived from Tensor. |
C++: lcn_layer_dense_batch.cpp
Java*: LCNLayerDenseBatch.java
Python*: lcn_layer_dense_batch.py