brainpy.layers module

Basic ANN Layer Class

Layer([name, mode])

Base class for a layer of artificial neural network.

Convolutional Layers

Conv1d(in_channels, out_channels, kernel_size)	One-dimensional convolution.
Conv2d(in_channels, out_channels, kernel_size)	Two-dimensional convolution.
Conv3d(in_channels, out_channels, kernel_size)	Three-dimensional convolution.

Dropout Layers

Dropout(prob[, seed, mode, name])

A layer that stochastically ignores a subset of inputs each training step.

Function Layers

Activation(activate_fun[, name, mode])	Applies an activation function to the inputs
Flatten([name, mode])	Flattens a contiguous range of dims into 2D or 1D.
FunAsLayer(fun[, name, mode, has_shared])

Dense Connection Layers

Dense(num_in, num_out[, W_initializer, ...])

A linear transformation applied over the last dimension of the input.

Normalization Layers

BatchNorm1d(num_features[, axis, epsilon, ...])	1-D batch normalization [1]_.
BatchNorm2d(num_features[, axis, epsilon, ...])	2-D batch normalization [1]_.
BatchNorm3d(num_features[, axis, epsilon, ...])	3-D batch normalization [1]_.
LayerNorm(normalized_shape[, epsilon, ...])	Layer normalization (https://arxiv.org/abs/1607.06450).
GroupNorm(num_groups, num_channels[, ...])	Group normalization layer.
InstanceNorm(num_channels[, epsilon, ...])	Instance normalization layer.

NVAR Layers

NVAR(num_in, delay[, order, stride, ...])

Nonlinear vector auto-regression (NVAR) node.

Pooling Layers

MaxPool(kernel_size[, stride, padding, ...])	Pools the input by taking the maximum over a window.
MinPool(kernel_size[, stride, padding, ...])	Pools the input by taking the minimum over a window.
AvgPool(kernel_size[, stride, padding, ...])	Pools the input by taking the average over a window.
AvgPool1d(kernel_size[, stride, padding, ...])	Applies a 1D average pooling over an input signal composed of several input
AvgPool2d(kernel_size[, stride, padding, ...])	Applies a 2D average pooling over an input signal composed of several input
AvgPool3d(kernel_size[, stride, padding, ...])	Applies a 3D average pooling over an input signal composed of several input
MaxPool1d(kernel_size[, stride, padding, ...])	Applies a 1D max pooling over an input signal composed of several input
MaxPool2d(kernel_size[, stride, padding, ...])	Applies a 1D max pooling over an input signal composed of several input
MaxPool3d(kernel_size[, stride, padding, ...])	Applies a 1D max pooling over an input signal composed of several input
AdaptiveAvgPool1d(target_shape[, ...])	Adaptive one-dimensional average down-sampling.
AdaptiveAvgPool2d(target_shape[, ...])	Adaptive two-dimensional average down-sampling.
AdaptiveAvgPool3d(target_shape[, ...])	Adaptive three-dimensional average down-sampling.
AdaptiveMaxPool1d(target_shape[, ...])	Adaptive one-dimensional maximum down-sampling.
AdaptiveMaxPool2d(target_shape[, ...])	Adaptive two-dimensional maximum down-sampling.
AdaptiveMaxPool3d(target_shape[, ...])	Adaptive three-dimensional maximum down-sampling.

Reservoir Layers

Reservoir(input_shape, num_out[, ...])

Reservoir node, a pool of leaky-integrator neurons with random recurrent connections [1]_.

Artificial Recurrent Layers

RNNCell(num_in, num_out[, ...])	Basic fully-connected RNN core.
GRUCell(num_in, num_out[, Wi_initializer, ...])	Gated Recurrent Unit.
LSTMCell(num_in, num_out[, Wi_initializer, ...])	Long short-term memory (LSTM) RNN core.
VanillaRNN(*args, **kwargs)	Vanilla RNN.
GRU(*args, **kwargs)	GRU.
LSTM(*args, **kwargs)	LSTM.