Layer parameter types

These structure types are used for parameters to the functions that create layers and filters.

Topics

Layer data

Describes a piece of training data such as weights or biases - a pointer to the raw data and information on the storage data type.

BNNSLayerData

A structure containing common layer parameters.

Activation

This structure specifies one of the enumerated activation functions and two parameter values to be passed to it.

BNNSActivationFunction

A structure containing layer functions.

BNNSActivation

A structure containing common activation function parameters.

Convolution layer parameters

Data specific to a convolution layer, including dimensions, number of channels, weights, biases, and activation function.

BNNSConvolutionLayerParameters

A structure containing convolution parameters.

Fully connected layer parameters

Data specific to a fully connected filter, including input, output, weights, biases, and activation function.

BNNSFullyConnectedLayerParameters

A structure containing fully connected layer parameters.

Pooling function

This enumeration defines the built-in pooling functions available in BNNS.

BNNSPoolingFunction

A structure containing pooling functions.

Pooling layer parameters

Data specific to a pooling layer, including dimensions, number of channels, weights, biases, and activation function.

BNNSPoolingLayerParameters

A structure containing pooling layer parameters.

See Also

BNNS Symbols

Common data types

This enumeration defines the basic data numeric types that can be specified in parameters to BNNS functions.

Function types

These are types for user-defined memory management functions.

Filter parameters type

These structure types are used for parameters to the functions that create layers and filters.

Create and destroy filters and layers

These functions are used to create and destroy layers with filters installed.

Apply filters

These functions apply a filter to an input or to a set of inputs.