q3.py

# -*- coding: utf-8 -*-
"""
Question 3:
Find the optimal number of hidden neurons for the 3-layer network designed.
a) Plot the training errors against number of epochs for the 3-layer network 
for different hidden-layer neurons. Limit search space to:{20,30,40,50,60}.
b) Plot the test errors against number of epochs for the optimum number of 
hidden layer neurons.
c) State the rationale behind selecting the optimal number of hidden neurons
"""

import numpy as np

from housing_preprocessing.preprocess import Preprocessor
from housing_training.approximation import Approximation
from housing_visualization.visualization import plot_graph


def main():
    print("Start Question 3")
    np.random.seed(10)

    k_fold = 5
    hidden_neuron_list = [[20], [30], [40], [50], [60]]
    batch_size = 32
    learning_rate = 1e-4
    epochs = 1000
    data_dir = "housing_data/cal_housing.data"

    # Preprocessing: Load data
    preprocessor = Preprocessor()
    preprocessor.load_data(data_dir)
    preprocessor.divide_data(3, 7)
    preprocessor.normalize_data()

    list_train_cost = []
    list_test_cost = []
    list_test_accuracy = []
    nn = Approximation()

    for hidden_neurons in hidden_neuron_list:
        if k_fold > 0:
            nn.select_model(
                train_x = preprocessor.train_x,
                train_y = preprocessor.train_y,
                k_fold = k_fold,
                epochs = epochs,
                batch_size = batch_size,
                hidden_neurons = hidden_neurons,
                learning_rate = learning_rate
            )

        else:
            nn.set_x_train(preprocessor.train_x)
            nn.set_y_train(preprocessor.train_y)
        
        nn.set_x_test(preprocessor.test_x)
        nn.set_y_test(preprocessor.test_y)
        nn.create_model(hidden_neurons, learning_rate)
        train_cost, test_cost, test_accuracy, min_err = nn.train_model(epochs=epochs, batch_size=batch_size, verbose=True)
        list_train_cost.append(train_cost)
        list_test_cost.append(test_cost)
        list_test_accuracy.append(test_accuracy)

    # Plot training error against number of epoch
    plot_graph(
        title='Train Errors for each Neuron',
        x_label="Epochs",
        y_label="MSE",
        x_val=range(epochs),
        y_vals=list_train_cost,
        data_labels=hidden_neuron_list,
    )

    # Plot test error of prediction against number of epoch
    plot_graph(
        title='Test Errors for each Neuron',
        x_label="Epochs",
        y_label="MSE",
        x_val=range(epochs),
        y_vals=list_test_cost,
        data_labels=hidden_neuron_list,
    )

    # Plot accuracy against number of epoch
    plot_graph(
        title="Test Accuracy",
        x_label="Epochs",
        y_label="Accuracy",
        x_val=range(epochs),
        y_vals=list_test_accuracy,
        data_labels=hidden_neuron_list,
    )

if __name__ == '__main__':
    main()