def red_wine_run(train_red_x, train_red_y, test_red_x, test_red_y):
# Red wine data
print('---------------\nRed Wine Data\n---------------')
# Training Phase
# values for 2D-grid search
lam = [] # regularization weight [min, max]
alpha = [] # learning rate [min, max]
nepochs = [] # sample # of epochs
epsilon = 0.0 # epsilon value
param = []
# end TODO
# using this alpha and lambda values run the training
print(f"alpha: {alpha}, lambda:{lam}")
print("Running Training phase")
# return param and optimal values for alpha and lambda from SGDSolver
param, alpha, lam = SGDSolver('Training', train_red_x, train_red_y, alpha,
lam, nepochs, epsilon, param)
# optimal values from 2-D search
print(f"optimal values\nalpha: {alpha}, lambda: {lam}")
# Note: validation and testing phases only take a single value for (alpha, lam) and not a list.
# Validation Phase
x_mse_val = SGDSolver('Validation', test_red_x, test_red_y, alpha, lam,
nepochs, epsilon, param)
print(f"Current Red Wine Data MSE is: {mse_val}.")
# Testing Phase
red_wine_predicted = SGDSolver('Testing', test_red_x, test_red_y, alpha,
lam, nepochs, epsilon, param)
for i in range(0, 50):
print(f"Predicted: {red_wine_predicted[i]}, Real: {test_red_y[i]}")
def white_wine_run(train_white_x, train_white_y, test_white_x, test_white_y):
# White wine data
print('---------------\nWhite Wine Data\n---------------')
# TODO: Change hyperparameter values here as needed
# similar to red_wine_run
# values for 2D-grid search
lam = [1e-3, 2] # regularization weight [min, max]
alpha = [1e-6, 1e-5] # learning rate [min, max]
nepochs = 500 # sample # of epochs
epsilon = 0.05 # epsilon value
param = np.random.standard_normal(size=np.shape(train_white_x)[1] + 1)
param = np.reshape(param, newshape=(
np.shape(train_white_x)[1] + 1,
1,
))
# end TODO
# Training Phase
print(f"alpha: {alpha}, lambda:{lam}")
print("Running Training phase")
# return param and optimal values for alpha and lambda from SGDSolver
training_start = time.time()
param, alpha, lam = SGDSolver('Training', train_white_x, train_white_y,
alpha, lam, nepochs, epsilon, param)
training_time = time.time() - training_start
# optimal values from 2-D search
print(f"optimal values\nalpha: {alpha}, lambda: {lam}")
# Note: validation and testing phases only take a single value for (alpha, lam) and not a list.
# Validation Phase
x_mse_val = SGDSolver('Validation', test_white_x, test_white_y, alpha, lam,
nepochs, epsilon, param)
print(f"Current White Wine Data MSE is: {x_mse_val}.")
# Testing Phase
test_start = time.time()
white_wine_predicted = SGDSolver('Testing', test_white_x, test_white_y,
alpha, lam, nepochs, epsilon, param)
test_time = time.time() - test_start
#for i in range(100, 150):
# print(f"Predicted: {white_wine_predicted[i]}, Real: {test_white_y[i]}")
print(
f'Training took: {training_time:.2f}\nTesting took: {test_time:.3f}s')
test_white_y = bc.classify_real_result(test_white_y)
confusion_matrix = cf.calculate_confusion_matrix(white_wine_predicted,
test_white_y)
accuracy = cf.calculate_accuracy(white_wine_predicted, test_white_y)
cf.print_confusion_matrix(confusion_matrix)
cf.print_accuracy(accuracy)
def white_wine_run(train_white_x, train_white_y, test_white_x, test_white_y):
# White wine data
print('---------------\nWhite Wine Data\n---------------')
# TODO: Change hyperparameter values here as needed
# similar to red_wine_run
# values for 2D-grid search
lam = [0.001, 0.01] # regularization weight [min, max]
alpha = [0.0, 0.1] # learning rate [min, max]
nepochs = 5 # sample # of epochs
epsilon = 0.0 # epsilon value
param = []
# end TODO
# Training Phase
print(f"alpha: {alpha}, lambda:{lam}")
print("Running Training phase")
# return param and optimal values for alpha and lambda from SGDSolver
param, alpha, lam = SGDSolver('Training', train_white_x, train_white_y,
alpha, lam, nepochs, epsilon, param)
# optimal values from 2-D search
print(f"optimal values\nalpha: {alpha}, lambda: {lam}")
# Note: validation and testing phases only take a single value for (alpha, lam) and not a list.
# Validation Phase
x_mse = SGDSolver('Validation', test_white_x, test_white_y, alpha, lam,
nepochs, epsilon, param)
print(f"Current White Wine Data MSE is: {x_mse}.")
print("\nAccuracy is: ")
acc = 1 - math.sqrt(x_mse)
print(acc)
# Testing Phase
white_wine_predicted = SGDSolver('Testing', test_white_x, test_white_y,
alpha, lam, nepochs, epsilon, param)
for i in range(0, 50):
print(f"Predicted: {white_wine_predicted[i]}, Real: {test_white_y[i]}")