Package org.deeplearning4j.nn.conf.layers

Class Summary

Class	Description
AbstractLSTM	LSTM recurrent net, based on Graves: Supervised Sequence Labelling with Recurrent Neural Networks http://www.cs.toronto.edu/~graves/phd.pdf
AbstractLSTM.Builder<T extends AbstractLSTM.Builder<T>>
ActivationLayer	Activation layer is a simple layer that applies the specified activation function to the input activations
ActivationLayer.Builder
AutoEncoder	Autoencoder layer.
AutoEncoder.Builder
BaseLayer	A neural network layer.
BaseLayer.Builder<T extends BaseLayer.Builder<T>>
BaseOutputLayer
BaseOutputLayer.Builder<T extends BaseOutputLayer.Builder<T>>
BasePretrainNetwork
BasePretrainNetwork.Builder<T extends BasePretrainNetwork.Builder<T>>
BaseRecurrentLayer
BaseRecurrentLayer.Builder<T extends BaseRecurrentLayer.Builder<T>>
BaseUpsamplingLayer	Upsampling base layer
BaseUpsamplingLayer.UpsamplingBuilder<T extends BaseUpsamplingLayer.UpsamplingBuilder<T>>
BatchNormalization	Batch normalization layer See: Ioffe and Szegedy, 2014, Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift https://arxiv.org/abs/1502.03167
BatchNormalization.Builder
CapsuleLayer	An implementation of the DigiCaps layer from Dynamic Routing Between Capsules Input should come from a PrimaryCapsules layer and be of shape [mb, inputCaps, inputCapDims].
CapsuleLayer.Builder
CapsuleStrengthLayer	An layer to get the "strength" of each capsule, that is, the probability of it being in the input.
CapsuleStrengthLayer.Builder
CenterLossOutputLayer	Center loss is similar to triplet loss except that it enforces intraclass consistency and doesn't require feed forward of multiple examples.
CenterLossOutputLayer.Builder
Cnn3DLossLayer	3D Convolutional Neural Network Loss Layer. Handles calculation of gradients etc for various loss (objective) functions. NOTE: Cnn3DLossLayer does not have any parameters.
Cnn3DLossLayer.Builder
CnnLossLayer	Convolutional Neural Network Loss Layer. Handles calculation of gradients etc for various loss (objective) functions. NOTE: CnnLossLayer does not have any parameters.
CnnLossLayer.Builder
Convolution1D	1D convolution layer.
Convolution1DLayer	1D (temporal) convolutional layer.
Convolution1DLayer.Builder
Convolution2D	2D convolution layer
Convolution3D	3D convolution layer configuration
Convolution3D.Builder
ConvolutionLayer	2D Convolution layer (for example, spatial convolution over images).
ConvolutionLayer.BaseConvBuilder<T extends ConvolutionLayer.BaseConvBuilder<T>>
ConvolutionLayer.Builder
Deconvolution2D	2D deconvolution layer configuration Deconvolutions are also known as transpose convolutions or fractionally strided convolutions.
Deconvolution2D.Builder
DenseLayer	Dense layer: a standard fully connected feed forward layer
DenseLayer.Builder
DepthwiseConvolution2D	2D depth-wise convolution layer configuration.
DepthwiseConvolution2D.Builder
DropoutLayer	Dropout layer.
DropoutLayer.Builder
EmbeddingLayer	Embedding layer: feed-forward layer that expects single integers per example as input (class numbers, in range 0 to numClass-1) as input.
EmbeddingLayer.Builder
EmbeddingSequenceLayer	Embedding layer for sequences: feed-forward layer that expects fixed-length number (inputLength) of integers/indices per example as input, ranged from 0 to numClasses - 1.
EmbeddingSequenceLayer.Builder
FeedForwardLayer	Created by jeffreytang on 7/21/15.
FeedForwardLayer.Builder<T extends FeedForwardLayer.Builder<T>>
GlobalPoolingLayer	Global pooling layer - used to do pooling over time for RNNs, and 2d pooling for CNNs. Supports the following PoolingTypes: SUM, AVG, MAX, PNORM Global pooling layer can also handle mask arrays when dealing with variable length inputs.
GlobalPoolingLayer.Builder
GravesBidirectionalLSTM	Deprecated use Bidirectional instead.
GravesBidirectionalLSTM.Builder
GravesLSTM	Deprecated Will be eventually removed.
GravesLSTM.Builder
InputTypeUtil	Utilities for calculating input types
Layer	A neural network layer.
Layer.Builder<T extends Layer.Builder<T>>
LayerValidation	Utility methods for validating layer configurations
LearnedSelfAttentionLayer	Implements Dot Product Self Attention with learned queries Takes in RNN style input in the shape of [batchSize, features, timesteps] and applies dot product attention using learned queries.
LearnedSelfAttentionLayer.Builder
LocallyConnected1D	SameDiff version of a 1D locally connected layer.
LocallyConnected1D.Builder
LocallyConnected2D	SameDiff version of a 2D locally connected layer.
LocallyConnected2D.Builder
LocalResponseNormalization	Local response normalization layer See section 3.3 of http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf
LocalResponseNormalization.Builder
LossLayer	LossLayer is a flexible output layer that performs a loss function on an input without MLP logic. LossLayer is similar to OutputLayer in that both perform loss calculations for network outputs vs.
LossLayer.Builder
LSTM	LSTM recurrent neural network layer without peephole connections.
LSTM.Builder
NoParamLayer
OutputLayer	Output layer used for training via backpropagation based on labels and a specified loss function.
OutputLayer.Builder
Pooling1D	1D Pooling (subsampling) layer.
Pooling2D	2D Pooling (subsampling) layer.
PReLULayer	Parametrized Rectified Linear Unit (PReLU)
PReLULayer.Builder
PrimaryCapsules	An implementation of the PrimaryCaps layer from Dynamic Routing Between Capsules Is a reshaped 2D convolution, and the input should be 2D convolutional ([mb, c, h, w]).
PrimaryCapsules.Builder
RecurrentAttentionLayer	Implements Recurrent Dot Product Attention Takes in RNN style input in the shape of [batchSize, features, timesteps] and applies dot product attention using the hidden state as the query and all time steps as keys/values.
RecurrentAttentionLayer.Builder
RnnLossLayer	Recurrent Neural Network Loss Layer. Handles calculation of gradients etc for various objective (loss) functions. Note: Unlike RnnOutputLayer this RnnLossLayer does not have any parameters - i.e., there is no time distributed dense component here.
RnnLossLayer.Builder
RnnOutputLayer	A version of OutputLayer for recurrent neural networks.
RnnOutputLayer.Builder
SelfAttentionLayer	Implements Dot Product Self Attention Takes in RNN style input in the shape of [batchSize, features, timesteps] and applies dot product attention using each timestep as the query.
SelfAttentionLayer.Builder
SeparableConvolution2D	2D Separable convolution layer configuration.
SeparableConvolution2D.Builder
SpaceToBatchLayer	Space to batch utility layer configuration for convolutional input types.
SpaceToBatchLayer.Builder<T extends SpaceToBatchLayer.Builder<T>>
SpaceToDepthLayer	Space to channels utility layer configuration for convolutional input types.
SpaceToDepthLayer.Builder<T extends SpaceToDepthLayer.Builder<T>>
Subsampling1DLayer	1D (temporal) subsampling layer - also known as pooling layer. Expects input of shape [minibatch, nIn, sequenceLength].
Subsampling1DLayer.Builder
Subsampling3DLayer	3D subsampling / pooling layer for convolutional neural networks
Subsampling3DLayer.BaseSubsamplingBuilder<T extends Subsampling3DLayer.BaseSubsamplingBuilder<T>>
Subsampling3DLayer.Builder
SubsamplingLayer	Subsampling layer also referred to as pooling in convolution neural nets Supports the following pooling types: MAX, AVG, SUM, PNORM
SubsamplingLayer.BaseSubsamplingBuilder<T extends SubsamplingLayer.BaseSubsamplingBuilder<T>>
SubsamplingLayer.Builder
Upsampling1D	Upsampling 1D layer Repeats each step size times along the temporal/sequence axis (dimension 2) For input shape [minibatch, channels, sequenceLength] output has shape [minibatch, channels, size * sequenceLength] Example:
Upsampling1D.Builder
Upsampling2D	Upsampling 2D layer Repeats each value (or rather, set of depth values) in the height and width dimensions by size[0] and size[1] times respectively. If input has shape [minibatch, channels, height, width] then output has shape [minibatch, channels, height*size[0], width*size[1]] Example:
Upsampling2D.Builder
Upsampling3D	Upsampling 3D layer Repeats each value (all channel values for each x/y/z location) by size[0], size[1] and size[2] If input has shape [minibatch, channels, depth, height, width] then output has shape [minibatch, channels, size[0] * depth, size[1] * height, size[2] * width]
Upsampling3D.Builder
ZeroPadding1DLayer	Zero padding 1D layer for convolutional neural networks.
ZeroPadding1DLayer.Builder
ZeroPadding3DLayer	Zero padding 3D layer for convolutional neural networks.
ZeroPadding3DLayer.Builder
ZeroPaddingLayer	Zero padding layer for convolutional neural networks (2D CNNs).
ZeroPaddingLayer.Builder

Enum Summary

Enum	Description
Convolution3D.DataFormat	An optional dataFormat: "NDHWC" or "NCDHW".
ConvolutionLayer.AlgoMode	The "PREFER_FASTEST" mode will pick the fastest algorithm for the specified parameters from the ConvolutionLayer.FwdAlgo, ConvolutionLayer.BwdFilterAlgo, and ConvolutionLayer.BwdDataAlgo lists, but they may be very memory intensive, so if weird errors occur when using cuDNN, please try the "NO_WORKSPACE" mode.
ConvolutionLayer.BwdDataAlgo	The backward data algorithm to use when ConvolutionLayer.AlgoMode is set to "USER_SPECIFIED".
ConvolutionLayer.BwdFilterAlgo	The backward filter algorithm to use when ConvolutionLayer.AlgoMode is set to "USER_SPECIFIED".
ConvolutionLayer.FwdAlgo	The forward algorithm to use when ConvolutionLayer.AlgoMode is set to "USER_SPECIFIED".
PoolingType	Pooling type: MAX: Max pooling - output is the maximum value of the input values AVG: Average pooling - output is the average value of the input values SUM: Sum pooling - output is the sum of the input values PNORM: P-norm pooling
SpaceToDepthLayer.DataFormat
Subsampling3DLayer.PoolingType
SubsamplingLayer.PoolingType