generate_lstm_model

Generate LSTM model

This module implements functions to generate a LSTM model for trajectory classification.

generate_lstm_model.generate_lstm_model(sim_df, parms)

Prepares the dataset, defines the neural network, trains and evaluates the model. Best model is automatically saved.

Parameters:

• sim_df (pd.DataFrame) – Dataframe containing the simulated tracks for the training.

• parms (dict) – Stored parameters containing global variables and instructions.

generate_lstm_model.plot_accuracy_over_window(path, booleans, window_size)

Plots a bar graph of the accuracy gradient over the tracks.

Parameters:

• path (str) – Output path.

• booleans (np.array) – Array of true and false based on ‘predicted_states_test == true_states_test’.

• window_size (int) – Used window size.

generate_lstm_model.plot_training_curves(path, history, patience)

Plots loss and accuracy training curves.

Parameters:

• path (str) – Output path.

• history (History keras object) – History of the training to be plotted.

• patience (int) – Value of the EarlyStopping criterion used here to trim the axis.