RutherfordNet

class pyeem.analysis.models.RutherfordNet(**kwargs)

Bases: object

The convolutional neural network (CNN) described in Rutherford et al. 2020.

Methods Summary

create_model([name, input_shape, …])

Builds and compiles the CNN.

get_test_data(dataset, routine_results_df)

Assembles the test data in a format that is able to be ingested by the Keras CNN model.

get_training_data(dataset, ss_results_df, …)

Assembles a training data in a format that is able to be ingested by the Keras CNN model.

prepare_data(dataset, ss_results_df, …)

Assembles both training and test data in a format that is able to be ingested by the Keras CNN model.

train(X, y[, fit_kws])

Train the CNN model with a call to Keras’ fit().

Methods Documentation

create_model(name='rutherfordnet', input_shape=[142, 139, 1], output_dense_units=3, compile_kws={})

Builds and compiles the CNN.

Parameters

  • name (str, optional) – The name of the model. Defaults to “rutherfordnet”.

  • compile_kws (dict, optional) – Additional keyword arguments which will be passed to tensorflow.keras.Model.compile(). Defaults to {}.

Returns

The compiled CNN model.

Return type

tensorflow.keras.Model

get_test_data(dataset, routine_results_df)

Assembles the test data in a format that is able to be ingested by the Keras CNN model. This data will be fed into the trained CNN for it to make predictions with.

Parameters

  • dataset (pyeem.datasets.Dataset) – The PyEEM dataset being used to generate test data.

  • routine_results_df (pandas.DataFrame) – The results of the preprocessing routine.

Returns

The formatted test data to be used in pyeem.analysis.models.RutherfordNet.model.predict()

Return type

tuple of numpy.ndarray

get_training_data(dataset, ss_results_df, mix_results_df)

Assembles a training data in a format that is able to be ingested by the Keras CNN model.

Parameters

  • dataset (pyeem.datasets.Dataset) – The PyEEM dataset being used to generate training data.

  • ss_results_df (pandas.DataFrame) – The augmented single source spectra results.

  • mix_results_df (pandas.DataFrame) – The augmented mixture spectra results.

Returns

The formatted training data to be used in pyeem.analysis.models.RutherfordNet.train()

Return type

tuple of numpy.ndarray

prepare_data(dataset, ss_results_df, mix_results_df, routine_results_df)

Assembles both training and test data in a format that is able to be ingested by the Keras CNN model.

Parameters

  • dataset (pyeem.datasets.Dataset) – A PyEEM dataset

  • ss_results_df (pandas.DataFrame) – The augmented single source spectra results.

  • mix_results_df (pandas.DataFrame) – The augmented mixture spectra results.

  • routine_results_df (pandas.DataFrame) – The results of the preprocessing routine.

Returns

Training and test data.

Return type

tuple of (tuple of numpy.ndarray)

train(X, y, fit_kws={})

Train the CNN model with a call to Keras’ fit().

Parameters

  • X (numpy.ndarray) – Training Spectra.

  • y (numpy.ndarray) – Concentration labels.

  • fit_kws (dict, optional) – Additional key word arguments which will be used in the call to Kera’s fit(). Defaults to {}.

Returns

The model’s training history which contains information about model accuracy and loss across training epochs.

Return type

tensorflow.python.keras.callbacks.History