dig.auggraph.method

GraphAug

An augmentation method for graph datasets under dig.auggraph.method.GraphAug implemented from the paper Automated Data Augmentations for Graph Classification.

RunnerAugCls	Runs the training of a graph classification model using the augmented data generated by the already trained generator.
RunnerGenerator	Runs the training of an augmented samples generator model which uses the already trained reward generation model.
RunnerRewardGen	Runs the training of a reward generation model which will be able to distinguish between graphs with different labels.

class RunnerAugCls(data_root_path, dataset_name, conf)

Runs the training of a graph classification model using the augmented data generated by the already trained generator. Check examples.auggraph.GraphAug.run_aug_cls for examples on how to run this augmented classifier model.

Parameters

• data_root_path (string) – Directory where datasets should be saved.

• dataset_name (dig.auggraph.method.GraphAug.constants.enums.DatasetName) – Name of the graph dataset.

• conf (dict) – Hyperparameters for the model. Check examples.auggraph.GraphAug.conf.aug_cls_conf for examples on how to define the conf dictionary for the augmented classifier model.

train_test(out_root_path, log_file='record.txt')

This method is used to run the training for the classification model on the augmented graph dataset and then validate the epoch parameters.

Parameters

• out_root_path (string) – Directory where the results of this augmented classifier model will be saved.

• log_file (string) – File where training and validation logs are written.

class RunnerGenerator(data_root_path, dataset_name, conf)

Runs the training of an augmented samples generator model which uses the already trained reward generation model. For a given graph, the model generates an augmented sample and a likelihood that this is a label invariant augmentation. This prediction is then evaluated by the reward generation model and a loss is computed based on these metrics. The loss is then minimized through training. Check examples.auggraph.GraphAug.run_generator for examples on how to run the generator model.

Parameters

• data_root_path (string) – Directory where datasets should be saved.

• dataset_name (dig.auggraph.method.GraphAug.constants.enums.DatasetName) – Name of the graph dataset.

• conf (dict) – Hyperparameters for the model. Check examples.auggraph.GraphAug.conf.generator_conf for examples on how to define the conf dictionary for the generator.

train_test(results_path)

This method is used to run the training for the augmented samples generator and validate the epoch parameters.

Parameters

results_path (string) – Directory where the resulting optimal parameters of the generator model will be saved.

class RunnerRewardGen(data_root_path, dataset_name, conf)

Runs the training of a reward generation model which will be able to distinguish between graphs with different labels. Check examples.auggraph.GraphAug.run_reward_gen for examples on how to run the reward generation model.

Parameters

• data_root_path (string) – Directory where datasets should be saved.

• dataset_name (dig.auggraph.method.GraphAug.constants.enums.DatasetName) – Name of the graph dataset.

• conf (dict) – Hyperparameters for the model. Check examples.auggraph.GraphAug.conf.reward_gen_conf for examples on how to define the conf dictionary for the reward generator.

train_test(results_path, num_save=30)

This method is used to run the training for the reward generation model and validate the epoch parameters.

Parameters

• results_path (string) – Directory where the resulting optimal parameters of the reward generation model will be saved.

• num_save (int) – Number of final epochs for which model parameters will be saved.