Package io.github.kirstenali.deepj.nn

package io.github.kirstenali.deepj.nn

Neural networks composed from DeepJ tensors, layers, autograd and backend operations.

Related Packages

Package	Description
io.github.kirstenali.deepj.nn.transformer

Class	Description
Activation	Adapts DeepJ activations to the existing Layer contract.
Autoencoder	Composable encoder/decoder, sharing the same parameter and autograd system as every DeepJ layer.
BatchNorm	Channel-axis (axis 1) normalization, using backend reductions and persistent FLOAT32 running statistics.
Conv1D	Temporal cross-correlation for [batch, channels, length], sharing the 2D implementation.
Conv2D	Cross-correlation of [batch, channels, height, width] inputs, using gather and matrix multiplication.
Dropout	Inverted dropout.
EncoderAttention	Bidirectional multi-head attention with optional padding masks and backend-resident softmax.
Flatten	Flattens an inclusive axis range; defaults to preserving the first (batch) axis.
FNN
GlobalAveragePool	Averages every spatial axis of [batch, channels, ...], returning [batch, channels].
GRU	Batch-first GRU layer; compose layers with Sequential for stacked recurrence.
GRUCell	Reset-after GRU, with reset/update/candidate gate order and separate projection biases.
LSTM	Batch-first LSTM layer; compose layers with Sequential for stacked recurrence.
LSTMCell	LSTM with input/forget/candidate/output gate order and explicit cell memory.
Module	Additive layer base with recursive training mode and identity-deduplicated parameters.
NdSupervisedTraining	First-axis minibatching for images, sequences and matrices, using DeepJ's existing Trainer.
NetworkWeights	Versioned parameter/buffer archives for a separately constructed network.
Pool1D	Max/average pooling of [batch, channels, length] inputs.
Pool2D	Floor-mode max/average pooling of channel-first images.
Pool2D.Reduction
Recurrent	Batch-first recurrent sequence execution, with optional lengths and caller-owned initial/final state.
RecurrentCell	Shared input/state validation and DeepJ linear projections for recurrent cells.
RecurrentResult	Batch-first sequence output and the final state of a recurrent layer.
RecurrentState	Explicit caller-owned state.
Reshape	Reshapes the full tensor, with at most one inferred (-1) dimension.
RNN	Batch-first RNN layer; compose layers with Sequential for stacked recurrence.
RNNCell	Elman recurrent cell with tanh activation.
Sequential	Ordered composition of existing or new DeepJ layers.
TensorJoin	Backend copies plus a slicing autograd rule; never reads tensor values on the host.
TransformerEncoder	Stacked encoder for [batch, time, width] embeddings, with padding masks and masked mean pooling.
TransformerEncoderConfig	Pre-normalized encoder dimensions.
TransformerEncoderLayer	Pre-normalized, bidirectional encoder block reusing DeepJ linear layers and LayerNorm.
Window2D	Floor-mode spatial window with symmetric zero padding and optional dilation.