Intel® Math Kernel Library 2018 Developer Reference - C
Creates propagation operations for layers performing local response normalization across channels.
dnnError_t dnnLRNCreateForward_F32 (dnnPrimitive_t *pLrn, dnnPrimitiveAttributes_t attributes, const dnnLayout_t dataLayout, size_t kernelSie, float alpha, float beta, float k);
dnnError_t dnnLRNCreateBackward_F32 (dnnPrimitive_t *pLrn, dnnPrimitiveAttributes_t attributes, const dnnLayout_t diffLayout, const dnnLayout_t dataLayout, size_t kernelSize, float alpha, float beta, float k);
dnnError_t dnnLRNCreateForward_F64 (dnnPrimitive_t *pLrn, dnnPrimitiveAttributes_t attributes, const dnnLayout_t dataLayout, size_t kernelSize, double alpha, double beta, double k);
dnnError_t dnnLRNCreateBackward_F64 (dnnPrimitive_t *pLrn, dnnPrimitiveAttributes_t attributes, const dnnLayout_t diffLayout, const dnnLayout_t dataLayout, size_t kernelSize, double alpha, double beta, double k);
attributes |
The set of attributes for the primitive. |
dataLayout |
The layout of the input. |
diffLayout |
The layout of the destination diff. |
kernelSize |
The number of channels to normalize across. |
alpha |
The α parameter. |
beta |
The β parameter. |
k |
The k parameter. |
pLrn |
Pointer to the primitive to create:
|
Each dnnLRNCreate function creates a forward or backward propagation operation for local response normalization (LRN) across channels. The LRN operation is defined as: