npdl.layers

• Base Layers
• Core Layers
• Convolution Layers
• Embedding Layer
• Normalization Layer
• Pooling Layers
• Recurrent Layers
• Shape Layers

Base Layers

Layer

The Layer class represents a single layer of a neural network.

Core Layers

Linear	A fully connected layer implemented as the dot product of inputs and weights.
Dense	A fully connected layer implemented as the dot product of inputs and weights.
Softmax	A fully connected layer implemented as the dot product of inputs and weights.
Dropout	A dropout layer.

Convolution Layers

Convolution

Convolution operator for filtering windows of two-dimensional inputs.

Embedding Layer

Embedding

Normalization Layer

BatchNormal

Batch normalization layer (Ioffe and Szegedy, 2014) [R1] .

Pooling Layers

MeanPooling	Average pooling operation for spatial data.
MaxPooling	Max pooling operation for spatial data.

Recurrent Layers

Recurrent	A recurrent neural network (RNN) is a class of artificial neural network where connections between units form a directed cycle.
SimpleRNN	Fully-connected RNN where the output is to be fed back to input.
GRU	Gated recurrent units (GRUs) are a gating mechanism in recurrent neural networks, introduced in 2014.
LSTM	Long short-term memory (LSTM) is a recurrent neural network (RNN) architecture (an artificial neural network) proposed in 1997 by Sepp Hochreiter and Jürgen Schmidhuber [R7] and further improved in 2000 by Felix Gers et al.[R8]_ Like most RNNs, a LSTM network is universal in the sense that given enough network units it can compute anything a conventional computer can compute, provided it has the proper weight matrix, which may be viewed as its program.
BatchLSTM	Long short-term memory (LSTM) is a special kind of RNN.

Shape Layers

Flatten