C++ API Reference for Intel® Data Analytics Acceleration Library 2018 Update 3

References | Namespaces | Classes | Enumerations
Two-dimensional Convolution Layer

Contains classes for neural network 2D convolution layer. More...

References

 Backward Two-dimensional Convolution Layer
 Contains classes for the backward 2D convolution layer.
 
 Forward Two-dimensional Convolution Layer
 Contains classes for the forward 2D convolution layer.
 

Namespaces

 daal::algorithms::neural_networks::layers::convolution2d
 Contains classes for neural network 2D convolution layer.
 
 daal::algorithms::neural_networks::layers::convolution2d::interface1
 Contains version 1.0 of Intel(R) Data Analytics Acceleration Library (Intel(R) DAAL) interface.
 

Classes

class  Batch< algorithmFPType, method >
 Computes the result of the forward and backward 2D convolution layer of neural network in the batch processing mode. More...
 
struct  KernelSizes
 Data structure representing the size of the two-dimensional kernel subtensor. More...
 
struct  Strides
 Data structure representing the intervals on which the subtensors for 2D convolution are selected. More...
 
struct  Paddings
 Data structure representing the number of data elements to implicitly add to each size of the two-dimensional subtensor on which 2D convolution is performed. More...
 
struct  Indices
 Data structure representing the indices of the two dimensions on which 2D convolution is performed. More...
 
class  Parameter
 2D convolution layer parameters More...
 

Enumerations

enum  Method { defaultDense = 0 }
 
enum  LayerDataId { auxData, auxWeights }
 

Enumeration Type Documentation

enum LayerDataId

Available identifiers of results of the forward 2D convolution layer and input objects for the backward 2D convolution layer

Enumerator
auxData 

Data processed at the forward stage of the layer

auxWeights 

Input weights for forward stage of the layer

enum Method

Available methods to compute forward and backward 2D convolution layer

Enumerator
defaultDense 

Default: performance-oriented method.

For more complete information about compiler optimizations, see our Optimization Notice.