tinybig.layer

RPN Layers

In this module, we implement several frequently used RPN layers that can be used to design and construct the RPN models.

The layers implemented in this module all inherit from the tinybig.module.base_layer.layer class. For most of the layer, they will only re-define the init method to provide the component functions. Meanwhile, for some of the layers, they will also implement some other methods to re-define the internal calculation process, which will be clearly specified in the class implementation.

Classes in this Module

This module contains the following categories of component layers:

• Basic layers
• Grid based layers
• Chain based layers
• Graph based layers
• Bilinear layers

---

Organization of this Module

RPN Layer

• base_layer (defined in the tinybig.module directory)

Basic Layers

• perceptron_layer

Grid based Layers

• grid_interdependence_layer
• grid_compression_layer

Chain based Layers

• chain_interdependence_layer

Graph based Layers

• graph_interdependence_layer
• graph_bilinear_interdependence_layer

Bilinear Layers

• bilinear_interdependence_layer