sketchgraphs_models.graph

This module contains the main implementation for the graph models.

The graph models represent a family of models which recursively build the sketch by viewing the sketch as a graph of entities and constraints. This model is configurable in the amount of continuous parameters it models. In particular, by using the --disable_entity_features or --disable_edge_features switches, one can disable the modelling of parameters attached to either edges or entities (or both). Modelling sketches without entity features can lead to meaningless sketches before solving, but the presence of constraints can often lead to reasonable solved states.

Modules

sketchgraphs_models.graph.dataset	Main preprocessing strategies for model that encompasses the “subnode” model plus both constraint and entity parameters.
sketchgraphs_models.graph.entropy_rate	Module to compute the entropy rate of a given representation.
sketchgraphs_models.graph.eval_likelihood	Module to evaluate the graph model according to data likelihood.
sketchgraphs_models.graph.model	Implementation of main graph-based model.
sketchgraphs_models.graph.sample	This module implements sampling from the entity model.
sketchgraphs_models.graph.train	Main module for training entity model.