Prediction Functions

Functions to build and use predictive Deep Neural Networks to find plasma pedestal heights.

predict.build_and_test_single(csv_name, input_cols, output_col, plot_col=None, learning_rate=None, epochs=None, batch_size=None, maxoutput=None)

Build and test a model predicting a single column of data from multiple input columns. Calculate the mean squared error of model predictions.

Parameters
  • csv_name (string) – filename of csv.

  • input_cols (list) – list of column names (as strings) of engineering parameters to predict from.

  • output_col (string) – name of column you’d like to predict.

  • plot_col (string) – optional. Name of column you’d like to plot your predictions against, selected from the engineering parameters.

  • learning_rate (integer) – optional. Value from 0.1 - 0.0001 representing the learning rate of the Keras Adam optimizer. Documentation for the Adam optimizer can be found here: https://keras.io/api/optimizers/adam/.

  • epochs (integer) – optional. Number of epochs for training the DNN. Typical starting values may be 50 for a smaller dataset up to 500 for a large dataset. If not given, the function will use a default value of 50.

  • batch_size (integer) – optional. Batch size for training the DNN, either 8, 16, 32 or 64. Smaller batch sizes may give more accurate predictions but take more time. If not given, the function will use a default value of 8.

  • maxoutput (Boolean) – optional. Set equal to True to display a more detailed output, useful for debugging.

Returns

pedestal height predictions. integer: mean squared error of these predictions. Tensorflow Model: trained DNN model.

Return type

array

predict.get_train_test_data(dataset)

Split a chosen dataset randomly into an 80/20 train-test split. Display Numpy shapes of train and test data.

Parameters

dataset (string) – filename of csv.

Returns

training data. array: test data.

Return type

array

predict.predict_single(model, filename, input_cols)

Predict pedestal heights using a trained DNN created with build_and_test_single.

Parameters
  • model (Tensorflow Model) – trained DNN from build_and_test_single to predict a column with.

  • input_cols (array) – input columns to apply the trained model on.

Returns

predicted pedestal heights.

Return type

array