Batch Processing

Layer Input

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:

  • 1/mnij of size n1 x n3 x n4 if sumDimension is not NULL
  • 1/mnqij of size n1 x n2 x n3 x n4 if sumDimension is NULL

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.

Layer Parameters

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.

Layer Output

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.

For more complete information about compiler optimizations, see our Optimization Notice.
Select sticky button color: 
Orange (only for download buttons)