def test_linreg():
'''
Helper function that tests LinearRegression.
@param:
None
@return:
None
'''
X_train, X_test, Y_train, Y_test = import_wine(WINE_FILE_PATH)
num_features = X_train.shape[1]
# Padding the inputs with a bias
X_train_b = np.append(X_train, np.ones((len(X_train), 1)), axis=1)
X_test_b = np.append(X_test, np.ones((len(X_test), 1)), axis=1)
#### Stochastic Gradient Descent ######
print('---------- LINEAR REGRESSION w/ SGD ----------')
sgd_model = LinearRegression(num_features, sgd=True)
sgd_model.train(X_train_b, Y_train)
print('Average Training Loss:', sgd_model.average_loss(X_train_b, Y_train))
print('Average Testing Loss:', sgd_model.average_loss(X_test_b, Y_test))
#### Matrix Inversion ######
print('---- LINEAR REGRESSION w/ Matrix Inversion ---')
solver_model = LinearRegression(num_features)
solver_model.train(X_train_b, Y_train)
print('Average Training Loss:',
solver_model.average_loss(X_train_b, Y_train))
print('Average Testing Loss:', solver_model.average_loss(X_test_b, Y_test))
def test_linreg():
'''
Helper function that tests LinearRegression.
@param:
None
@return:
None
'''
m = np.array([[2, 3], [1, 0]])
mm = np.array([[2, 3, 4, 5], [1, 1, 3, 0]])
for l in range(2):
print(mm[l, range(2)])
###print(m)
n = np.append(m, np.ones((len(m), 1)), axis=1)
#print(n)
#print(m.shape[1])
s = LinearRegression(m.shape[1])
#print(s.weights)
X_train, X_test, Y_train, Y_test = import_wine(WINE_FILE_PATH)
#print(X_train.shape[1])
num_features = X_train.shape[1]
# Padding the inputs with a bias
X_train_b = np.append(X_train, np.ones((len(X_train), 1)), axis=1)
X_test_b = np.append(X_test, np.ones((len(X_test), 1)), axis=1)
#print(6.7**2)
#### Matrix Inversion ######
print('---- LINEAR REGRESSION w/ Matrix Inversion ---')
solver_model = LinearRegression(num_features)
solver_model.train(X_train_b, Y_train)
print('Average Training Loss:',
solver_model.average_loss(X_train_b, Y_train))
print('Average Testing Loss:', solver_model.average_loss(X_test_b, Y_test))
def test_models(dataset, epochs, test_size=0.2):
'''
Tests LinearRegression, OneLayerNN, TwoLayerNN on a given dataset.
:param dataset The path to the dataset
:return None
'''
# Check if the file exists
if not os.path.exists(dataset):
print('The file {} does not exist'.format(dataset))
exit()
# Load in the dataset
data = np.loadtxt(dataset, skiprows=1)
X, Y = data[:, 1:], data[:, 0]
# Normalize the features
X = (X - np.mean(X, axis=0)) / np.std(X, axis=0)
X_train, X_test, Y_train, Y_test = train_test_split(X,
Y,
test_size=test_size)
print('Running models on {} dataset'.format(dataset))
#### Linear Regression ######
print('----- LINEAR REGRESSION -----')
# Add a bias
X_train_b = np.append(X_train, np.ones((len(X_train), 1)), axis=1)
X_test_b = np.append(X_test, np.ones((len(X_test), 1)), axis=1)
regmodel = LinearRegression()
regmodel.train(X_train_b, Y_train)
print('Average Training Loss:', regmodel.average_loss(X_train_b, Y_train))
print('Average Testing Loss:', regmodel.average_loss(X_test_b, Y_test))
#### 1-Layer NN ######
print('----- 1-Layer NN -----')
nnmodel = OneLayerNN()
nnmodel.train(X_train_b, Y_train, epochs=epochs, print_loss=False)
print('Average Training Loss:', nnmodel.average_loss(X_train_b, Y_train))
print('Average Testing Loss:', nnmodel.average_loss(X_test_b, Y_test))
#### 2-Layer NN ######
print('----- 2-Layer NN -----')
model = TwoLayerNN(5)
# Use X without a bias, since we learn a bias in the 2 layer NN.
model.train(X_train, Y_train, epochs=epochs, print_loss=False)
print('Average Training Loss:', model.average_loss(X_train, Y_train))
print('Average Testing Loss:', model.average_loss(X_test, Y_test))
# Add some noise to the observations
noise_var = 0.5
# Create random input and output data
X = lhs(D_in, N)
y = 5 * X + noise_var * np.random.randn(N, D_out)
# Define the model
model = LinearRegression(X, y)
# Define an optimizer
optimizer = SGD(model.num_params, lr=1e-3, momentum=0.9)
# optimizer = Adam(model.num_params, lr = 1e-3)
# optimizer = RMSprop(model.num_params, lr = 1e-3)
# Train the model
model.train(10000, optimizer)
# Print the learned parameters
print('w = %e, sigma_sq = %e' %
(model.theta[:-1], np.exp(model.theta[-1])))
# Make predictions
y_pred = model.predict(X)
# Plot
plt.figure(1)
plt.plot(X, y, 'o')
plt.plot(X, y_pred)
plt.show()
from sklearn.metrics import mean_squared_error, r2_score
import config_loader
window_size = config_loader.get_window_size()
num_days = config_loader.get_num_predicted_days()
# Load the training data with n days held out
print("Loading training data...")
X_train, y_train, y_test = DataLoader.get_date_separated_testing_data(
window_size,
num_days,
)
# train the model (and save it to file)
print("Training the model...")
model = LinearRegression()
model.train(X_train, y_train)
# get metrics for the multi-day predictions in each state
states = DataLoader.get_states()
all_errors = np.empty((0, num_days))
all_predictions = np.empty((0, num_days))
all_actual = np.empty((0, num_days))
all_control = np.empty((0, num_days))
for state_name, state_abbrev in states.items():
if state_abbrev != "US":
# get the multi-day predictions
case_df = DataLoader.get_daily_cases_df(state_abbrev)[:-num_days]
vax_df = DataLoader.get_daily_vaccinations_df(state_abbrev)[:-num_days]
future_vaccinations = DataLoader.get_assumed_vaccinations_dict(
vax_df, num_days, multiplier=1)
predictions_dict = DataLoader.get_predictions(
# Test scipt
lm = LinearRegression()
n = 50
# Generate some noisy train data
noise = np.random.randn(n)
x = np.linspace(0, 20, n)
y = 32.3 + 5.2 * (x + noise)
x = scaleFeature(x)
y = scaleFeature(y)
x = x.reshape((-1, 1))
# Train the lm, print out the parameters, plot the fit
lm.train(x, y, 0.5, 100, 0)
print(lm.parameters)
# Plot the fit of the linear model
y_hat = lm.predict(x)
# Plot the fit of line and train data
plt.plot(x, y, 'o')
plt.plot(x, y_hat)
plt.ylabel("y")
plt.xlabel("x")
plt.title("Plot of the Models Fit")
plt.show()
wait()