Assignment 4 Part 1 - Gradient Descent

gradient_descent_math.py

import math

dot_product = lambda X, Y: sum(map(lambda x, y: x * y, X, Y))

def sigma(t):
    return 1 / (1 + math.exp(-t))
    
def sigma_prime(t):
    sigma_value = sigma(t)
    return sigma_value * (1 - sigma_value)

main.py

import sys
from gradient_descent_math import dot_product, sigma, sigma_prime

initial_weight = 0

def read_training_file(training_file_name):
    training_file = open(training_file_name, 'r')
    attributes = []
    examples = []
    correct_values = []
    raw_data = training_file.read().splitlines()
    for i in range(len(raw_data)):
        line_split = raw_data[i].split('\t')
        if i == 0:
            for datum in line_split:
                attributes.append(datum)
        else:
            examples.append([int(line_split[x]) for x in range(len(line_split) - 1)])
            correct_values.append(int(line_split[len(line_split) - 1]))
    return (attributes, examples, correct_values)

def calculate_weights(weights, examples, correct_values, iteration, learning_rate):
    global dot_product
    
    example = examples[iteration]
    correct_value = correct_values[iteration]
    predicted_value = sigma(dot_product(weights, example))
    delta = correct_value - predicted_value
    predicted_value_slope = sigma_prime(dot_product(weights, example))
    
    correction = [x * learning_rate * delta * predicted_value_slope for x in example]
    return [sum(w) for w in zip(weights, correction)]

def print_iteration(attributes, weights, examples, iteration):
    example = examples[iteration]
    predicted_value = sigma(dot_product(weights, example))

    print('After iteration ' + str(iteration + 1) + ': ', end='')
    for i in range(len(weights)):
        print('w(' + attributes[i] + ') = %.4f, ' % weights[i], end='')
    print('output = %.4f' % predicted_value)

if len(sys.argv) != 5:
    print('The program must be run with the training file, the test file, the learning rate, and the number of iterations as input.')
else:
    training_file = sys.argv[1]
    test_file = sys.argv[2]
    learning_rate = float(sys.argv[3])
    iterations = int(sys.argv[4])
    
    attributes, examples, correct_values = read_training_file(training_file)
    weights = [initial_weight] * len(examples[0])
    
    for iteration in range(iterations):
        weights = calculate_weights(weights, examples, correct_values, iteration, learning_rate)
        print_iteration(attributes, weights, examples, iteration)