Modifier and Type  Method and Description 

TrainingConfig 
Trainable.getConfig() 
Modifier and Type  Class and Description 

class 
AttentionVertex
Implements Dot Product Attention using the given inputs.

Modifier and Type  Class and Description 

class 
AbstractLSTM
LSTM recurrent net, based on Graves: Supervised Sequence Labelling with Recurrent Neural Networks
http://www.cs.toronto.edu/~graves/phd.pdf

class 
ActivationLayer
Activation layer is a simple layer that applies the specified activation function to the input activations

class 
AutoEncoder
Autoencoder layer.

class 
BaseLayer
A neural network layer.

class 
BaseOutputLayer 
class 
BasePretrainNetwork 
class 
BaseRecurrentLayer 
class 
BaseUpsamplingLayer
Upsampling base layer

class 
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 
class 
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].

class 
CapsuleStrengthLayer
An layer to get the "strength" of each capsule, that is, the probability of it being in the input.

class 
CenterLossOutputLayer
Center loss is similar to triplet loss except that it enforces intraclass consistency and doesn't require feed
forward of multiple examples.

class 
Cnn3DLossLayer
3D Convolutional Neural Network Loss Layer.

class 
CnnLossLayer
Convolutional Neural Network Loss Layer.

class 
Convolution1D
1D convolution layer.

class 
Convolution1DLayer
1D (temporal) convolutional layer.

class 
Convolution2D
2D convolution layer

class 
Convolution3D
3D convolution layer configuration

class 
ConvolutionLayer
2D Convolution layer (for example, spatial convolution over images).

class 
Deconvolution2D
2D deconvolution layer configuration
Deconvolutions are also known as transpose convolutions or fractionally strided convolutions. 
class 
Deconvolution3D
3D deconvolution layer configuration
Deconvolutions are also known as transpose convolutions or fractionally strided convolutions. 
class 
DenseLayer
Dense layer: a standard fully connected feed forward layer

class 
DepthwiseConvolution2D
2D depthwise convolution layer configuration.

class 
DropoutLayer
Dropout layer.

class 
EmbeddingLayer
Embedding layer: feedforward layer that expects single integers per example as input (class numbers, in range 0 to
numClass1) as input.

class 
EmbeddingSequenceLayer
Embedding layer for sequences: feedforward layer that expects fixedlength number (inputLength) of integers/indices
per example as input, ranged from 0 to numClasses  1.

class 
FeedForwardLayer
Created by jeffreytang on 7/21/15.

class 
GlobalPoolingLayer
Global pooling layer  used to do pooling over time for RNNs, and 2d pooling for CNNs.

class 
GravesBidirectionalLSTM
Deprecated.
use
Bidirectional instead. With the Bidirectional
layer wrapper you can make any recurrent layer bidirectional, in particular GravesLSTM. Note that this layer adds the
output of both directions, which translates into "ADD" mode in Bidirectional.
Usage: .layer(new Bidirectional(Bidirectional.Mode.ADD, new GravesLSTM.Builder()....build())) 
class 
GravesLSTM
Deprecated.
Will be eventually removed. Use
LSTM instead, which has similar prediction accuracy, but supports
CuDNN for faster network training on CUDA (Nvidia) GPUs 
class 
Layer
A neural network layer.

class 
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.

class 
LocallyConnected1D
SameDiff version of a 1D locally connected layer.

class 
LocallyConnected2D
SameDiff version of a 2D locally connected layer.

class 
LocalResponseNormalization
Local response normalization layer
See section 3.3 of http://www.cs.toronto.edu/~fritz/absps/imagenet.pdf 
class 
LossLayer
LossLayer is a flexible output layer that performs a loss function on an input without MLP logic.

class 
LSTM
LSTM recurrent neural network layer without peephole connections.

class 
NoParamLayer 
class 
OutputLayer
Output layer used for training via backpropagation based on labels and a specified loss function.

class 
Pooling1D
1D Pooling (subsampling) layer.

class 
Pooling2D
2D Pooling (subsampling) layer.

class 
PReLULayer
Parametrized Rectified Linear Unit (PReLU)
f(x) = alpha * x for x < 0, f(x) = x for x >= 0
alpha has the same shape as x and is a learned parameter. 
class 
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]).

class 
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.

class 
RnnLossLayer
Recurrent Neural Network Loss Layer.

class 
RnnOutputLayer
A version of
OutputLayer for recurrent neural networks. 
class 
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.

class 
SeparableConvolution2D
2D Separable convolution layer configuration.

class 
SpaceToBatchLayer
Space to batch utility layer configuration for convolutional input types.

class 
SpaceToDepthLayer
Space to channels utility layer configuration for convolutional input types.

class 
Subsampling1DLayer
1D (temporal) subsampling layer  also known as pooling layer.

class 
Subsampling3DLayer
3D subsampling / pooling layer for convolutional neural networks
Supports max and average pooling modes

class 
SubsamplingLayer
Subsampling layer also referred to as pooling in convolution neural nets
Supports the following pooling types: MAX, AVG, SUM, PNORM

class 
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: If input (for a single example, with channels down page, and sequence from left to right) is: [ A1, A2, A3] [ B1, B2, B3] Then output with size = 2 is: [ A1, A1, A2, A2, A3, A3] [ B1, B1, B2, B2, B3, B2] 
class 
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. 
class 
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] 
class 
ZeroPadding1DLayer
Zero padding 1D layer for convolutional neural networks.

class 
ZeroPadding3DLayer
Zero padding 3D layer for convolutional neural networks.

class 
ZeroPaddingLayer
Zero padding layer for convolutional neural networks (2D CNNs).

Modifier and Type  Class and Description 

class 
Cropping1D
Cropping layer for convolutional (1d) neural networks.

class 
Cropping2D
Cropping layer for convolutional (2d) neural networks.

class 
Cropping3D
Cropping layer for convolutional (3d) neural networks.

Modifier and Type  Class and Description 

class 
ElementWiseMultiplicationLayer
Elementwise multiplication layer with weights: implements
out = activationFn(input .* w + b) where: w is a learnable weight vector of length nOut  ".*" is elementwise multiplication  b is a bias vector Note that the input and output sizes of the elementwise layer are the same for this layer created by jingshu 
class 
FrozenLayer
FrozenLayer is used for the purposes of transfer learning.

class 
FrozenLayerWithBackprop
Frozen layer freezes parameters of the layer it wraps, but allows the backpropagation to continue.

class 
RepeatVector
RepeatVector layer configuration.

Modifier and Type  Class and Description 

class 
Yolo2OutputLayer
Output (loss) layer for YOLOv2 object detection model, based on the papers: YOLO9000: Better, Faster, Stronger 
Redmon & Farhadi (2016)  https://arxiv.org/abs/1612.08242
and You Only Look Once: Unified, RealTime Object Detection  Redmon et al. (2016)  http://www.cvfoundation.org/openaccess/content_cvpr_2016/papers/Redmon_You_Only_Look_CVPR_2016_paper.pdf This loss function implementation is based on the YOLOv2 version of the paper. 
Modifier and Type  Class and Description 

class 
Bidirectional
Bidirectional is a "wrapper" layer: it wraps any unidirectional RNN layer to make it bidirectional.

class 
LastTimeStep
LastTimeStep is a "wrapper" layer: it wraps any RNN (or CNN1D) layer, and extracts out the last time step during forward pass,
and returns it as a row vector (per example).

class 
SimpleRnn
Simple RNN  aka "vanilla" RNN is the simplest type of recurrent neural network layer.

class 
TimeDistributed
TimeDistributed wrapper layer.

Modifier and Type  Class and Description 

class 
AbstractSameDiffLayer 
class 
SameDiffLambdaLayer
SameDiffLambdaLayer is defined to be used as the base class for implementing lambda layers using SameDiff
Lambda layers are layers without parameters  and as a result, have a much simpler API  users need only extend SameDiffLambdaLayer and implement a single method 
class 
SameDiffLambdaVertex
SameDiffLambdaVertex is defined to be used as the base class for implementing lambda vertices using SameDiff
Lambda vertices are vertices without parameters  and as a result, have a much simpler API  users need only extend SameDiffLambdaVertex and implement a single method to define their vertex 
class 
SameDiffLayer
A base layer used for implementing Deeplearning4j layers using SameDiff.

class 
SameDiffOutputLayer
A base layer used for implementing Deeplearning4j Output layers using SameDiff.

class 
SameDiffVertex
A SameDiffbased GraphVertex.

Modifier and Type  Class and Description 

class 
MaskLayer
MaskLayer applies the mask array to the forward pass activations, and backward pass gradients, passing through
this layer.

class 
MaskZeroLayer
Wrapper which masks timesteps with activation equal to the specified masking value (0.0 default).

Modifier and Type  Class and Description 

class 
VariationalAutoencoder
Variational Autoencoder layer
See: Kingma & Welling, 2013: AutoEncoding Variational Bayes  https://arxiv.org/abs/1312.6114
This implementation allows multiple encoder and decoder layers, the number and sizes of which can be set
independently.

Modifier and Type  Class and Description 

class 
BaseWrapperLayer
Base wrapper layer: the idea is to pass through all methods to the underlying layer, and selectively override
them as required.

Modifier and Type  Class and Description 

class 
DummyConfig
A 'dummy' training configuration for use in frozen layers

Modifier and Type  Class and Description 

class 
OCNNOutputLayer
An implementation of one class neural networks from:
https://arxiv.org/pdf/1802.06360.pdf
The one class neural network approach is an extension of the standard output layer with a single set of weights, an
activation function, and a bias to: 2 sets of weights, a learnable "r" parameter that is held static 1 traditional
set of weights. 1 additional weight matrix

Modifier and Type  Method and Description 

TrainingConfig 
BaseGraphVertex.getConfig() 
TrainingConfig 
BaseWrapperVertex.getConfig() 
Modifier and Type  Method and Description 

TrainingConfig 
FrozenVertex.getConfig() 
TrainingConfig 
LayerVertex.getConfig() 
Modifier and Type  Method and Description 

TrainingConfig 
AbstractLayer.getConfig() 
TrainingConfig 
FrozenLayer.getConfig() 
Modifier and Type  Method and Description 

TrainingConfig 
BidirectionalLayer.getConfig() 
Modifier and Type  Method and Description 

TrainingConfig 
SameDiffGraphVertex.getConfig() 
Modifier and Type  Class and Description 

class 
IdentityLayer
Identity layer, passes data through unaltered.

Modifier and Type  Method and Description 

TrainingConfig 
VariationalAutoencoder.getConfig() 
Modifier and Type  Method and Description 

TrainingConfig 
BaseWrapperLayer.getConfig() 
Modifier and Type  Class and Description 

class 
TFOpLayer 
Modifier and Type  Method and Description 

TrainingConfig 
MultiLayerNetwork.getConfig() 
Copyright © 2020. All rights reserved.