Layers

The athena library provides a variety of layers commonly used in building neural networks. These layers can be stacked and combined to create complex architectures for various machine learning tasks.

Available Layers

The following types of layers are available in the athena library:

• Core Layers

• Convolutional Layers

• Input Layers

• Merge Layers

• Message Passing Layers

• Neural Operator Layers

• Normalisation Layers

• Padding Layers

• Pooling Layers

• Recurrent Layers

• Regularisation Layers

• Reshaping Layers

Creating custom layers

The athena library is designed with extensibility in mind, allowing users to create custom layers by extending the base_layer_type.

Blocks

Blocks are a convenient way to group multiple layers together into a single unit that can be added to a network. An example of a custom block is the orthogonal_nop_block_type, which combines an orthogonal basis projection with a spectral mixing path and a local bypass.

See the tutorial: Creating Custom Layers