notebook module

class notebook.Notebook(config)

Bases: object

add_records_from(notebook_path)

get_top_policies(k)

Prints and returns top-k policies

Policies are ordered by their expected accuracy increas Args:

k (int) top-k

Returns

pandas.DataFrame: top-k policies as dataframe

output_top_policies()

record(trial_no, trial_hyperparams, sample_no, reward, history)

Records one complete training of child model

Args:

trial_no (int): no of trial (iteration) of training trial_hyperparams (list) : list of data augmentation hyperparameters used for training sample_no (int): sample no among training with same hyperparameters reward (float): reward is basically last n validation accuracy before overfitting history (dict): history returned by keras.model.fit()

save()

notebook.get_folder_path(path)