def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
params_n_h = [2, 8, 40, 100]
for n_h in params_n_h:
nn = MLPRegressor(solver='lbfgs',
max_iter=200,
activation='logistic',
hidden_layer_sizes=(n_h, ),
alpha=0,
verbose=False,
random_state=0)
# zero randomness
# verbose=True
nn.fit(x_train, y_train)
y_pred_train = nn.predict(x_train)
y_pred_test = nn.predict(x_test)
plot_learned_function(n_h, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
pass
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model.
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
nh = 50
nn = MLPRegressor(activation='logistic',
solver='lbfgs',
max_iter=5000,
alpha=0,
hidden_layer_sizes=(nh, ))
nn.fit(x_train, y_train)
y_pred_train = nn.predict(x_train)
y_pred_test = nn.predict(x_test)
plot_learned_function(nh, x_train, y_train, y_pred_train, x_test, y_test,
y_pred_test)
pass
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO - done
# using 8 hidden neurons
n_h = [2, 8, 40]
for i in n_h:
# use random state for a fixed split - guarantees always same output (200 seems beautiful)
nn = MLPRegressor(activation='logistic', solver='lbfgs', alpha=0.0, hidden_layer_sizes=(i,), max_iter=200,
random_state=200)
nn.fit(x_train, y_train)
pred_train_y = nn.predict(x_train)
pred_test_y = nn.predict(x_test)
plot_learned_function(n_h, x_train, y_train, pred_train_y, x_test, y_test, pred_test_y)
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
regressor = MLPRegressor(
hidden_layer_sizes=(40, ), #8,40
solver="lbfgs",
activation="logistic",
alpha=0.0,
max_iter=200,
)
regressor.fit(x_train, y_train)
#plot_learned_function(2,x_train, y_train,regressor.predict(x_train), x_test, y_test, regressor.predict(x_test))
#plot_learned_function(8, x_train, y_train, regressor.predict(x_train), x_test, y_test, regressor.predict(x_test))
plot_learned_function(40, x_train, y_train, regressor.predict(x_train),
x_test, y_test, regressor.predict(x_test))
pass
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
#pass
n_hidden = 5 # 2, 5, 50
reg = MLPRegressor(hidden_layer_sizes=(n_hidden, 8),
activation='logistic',
solver='lbfgs',
alpha=0)
reg.fit(x_train, y_train)
y_pred_test = reg.predict(x_test)
y_pred_train = reg.predict(x_train)
plot_learned_function(n_hidden, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
Use max_iter = 10000 and tol=1e-8
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
n_h = [1, 2, 3, 4, 6, 8, 12, 20, 40]
train_array = np.zeros((9, 10))
test_array = np.zeros((9, 10))
for n in n_h:
for i in range(0, 10):
index = n_h.index(n)
nn = MLPRegressor(tol=1e-8, activation='logistic', solver='lbfgs', alpha=0.0,
hidden_layer_sizes=(n_h[index],),
max_iter=10000, random_state=i)
nn.fit(x_train, y_train)
train_array[index][i] = calculate_mse(nn, x_train, y_train)
test_array[index][i] = calculate_mse(nn, x_test, y_test)
y_pred_train = nn.predict(x_train)
y_pred_test = nn.predict(x_test)
if n == 1:
plot_learned_function(n, x_train, y_train, y_pred_train, x_test, y_test, y_pred_test)
plot_mse_vs_neurons(np.array(train_array), np.array(test_array), n_h)
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
n_hidden = 40
regressor = MLPRegressor(hidden_layer_sizes=(n_hidden, ),
activation='logistic',
solver='lbfgs',
alpha=0,
max_iter=200)
regressor.fit(x_train, y_train)
y_pred_train = regressor.predict(x_train)
y_pred_test = regressor.predict(x_test)
plot_learned_function(n_hidden, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
# [train_mses, test_mses] = calculate_mse(regressor, [x_train, x_test], [y_train, y_test])
# plot_mse_vs_neurons(train_mses, test_mses, n_hidden_neurons_list)
pass
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
Use max_iter = 10000 and tol=1e-8
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
hidden_neurons_totest = np.array([1, 2, 3, 4, 6, 8, 12, 20, 40])
# hidden_neurons_totest = np.array([20])
dim1 = hidden_neurons_totest.shape[0]
mse_test_matrix = np.zeros((dim1, 10))
mse_train_matrix = np.zeros((dim1, 10))
k = 0
for i in hidden_neurons_totest:
n_hidden_neurons = i
for j in range(10):
nn = MLPRegressor(activation='logistic', solver='lbfgs', max_iter=10000, tol=1e-8,
hidden_layer_sizes=(n_hidden_neurons,), alpha=0, random_state=j)
nn.fit(x_train, y_train)
predictions_test = nn.predict(x_test)
mse_test_matrix[k, j] = calculate_mse(nn, x_test, y_test)
mse_train_matrix[k, j] = calculate_mse(nn, x_train, y_train)
k += 1
plot_mse_vs_neurons(mse_train_matrix, mse_test_matrix, hidden_neurons_totest)
plt.show()
plot_learned_function(40, x_train, y_train, 0, x_test, y_test, predictions_test)
plt.show()
def ex_1_1_d(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 b)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
#pass
N = 500
n_hidden = [2, 5, 50]
mse_train = np.zeros([np.size(n_hidden), N])
mse_test = np.zeros([np.size(n_hidden), N])
for j in range(np.size(n_hidden)):
reg = MLPRegressor(hidden_layer_sizes=(n_hidden[j], ),
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=0,
warm_start=True,
max_iter=1)
for r in range(N):
reg.fit(x_train, y_train)
mse_train[j, r] = calculate_mse(reg, x_train, y_train)
mse_test[j, r] = calculate_mse(reg, x_test, y_test)
# PLOT
plot_mse_vs_neurons(mse_train, mse_test, n_hidden)
#mse_test_mean = np.mean(mse_test, axis=1)
#ind = np.argmin(mse_test_mean)
ind = np.unravel_index(np.argmin(mse_test), mse_test.shape)
# geht es auch ohne den MLPRegressor nochmal zu initialisieren? Keine Ahnung obs anders besser geht
reg = MLPRegressor(hidden_layer_sizes=(n_hidden[j], ),
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=random.randint(0, 1000),
max_iter=500)
reg.fit(x_train, y_train)
y_pred_test = reg.predict(x_test)
y_pred_train = reg.predict(x_train)
plot_learned_function(n_hidden[ind[0]], x_train, y_train, y_pred_train,
x_test, y_test, y_pred_test)
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
Use max_iter = 10000 and tol=1e-8
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
m = 0
n = 0
#declaring variables used in MLP-Regressor
hidden_layers = np.array([1, 2, 3, 4, 6, 8, 12, 20, 40])
random_state = 10
activation_mode = 'logistic'
solver_mode = 'lbfgs'
alpha = 0
max_iter = 10000
tol = 1e-8
train_mse = np.zeros((hidden_layers.size, random_state))
test_mse = np.zeros((hidden_layers.size, random_state))
for m in range(random_state):
for n in range(hidden_layers.size):
nn = MLPRegressor(hidden_layer_sizes=(hidden_layers[n], ),
activation=activation_mode,
solver=solver_mode,
alpha=alpha,
max_iter=max_iter,
random_state=m,
tol=tol)
nn.fit(x_train, y_train)
#calculate for every random seed train_mse and test_mse
train_mse[n][m] = calculate_mse(nn, x_train, y_train)
test_mse[n][m] = calculate_mse(nn, x_test, y_test)
plot_mse_vs_neurons(train_mse, test_mse, hidden_layers)
y_test_pred = nn.predict(x_test)
y_train_pred = nn.predict(x_train)
plot_learned_function(40, x_train, y_train, y_train_pred, x_test, y_test,
y_test_pred)
pass
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
#pass
N = 10
n_hidden = [1, 2, 3, 4, 6, 8, 12, 20, 40]
mse_train = np.zeros([np.size(n_hidden), N])
mse_test = np.zeros([np.size(n_hidden), N])
for j in range(np.size(n_hidden)):
reg = MLPRegressor(hidden_layer_sizes=(1, n_hidden[j]),
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=random.randint(0, 1000))
for r in range(N):
reg.fit(x_train, y_train)
mse_train[j, r] = calculate_mse(reg, x_train, y_train)
mse_test[j, r] = calculate_mse(reg, x_test, y_test)
# PLOT
plot_mse_vs_neurons(mse_train, mse_test, n_hidden)
"""
mse_test_mean = np.mean(mse_test, axis=1)
ind = np.argmin(mse_test_mean)
"""
ind = np.unravel_index(np.argmin(mse_test), mse_test.shape)
reg = MLPRegressor(hidden_layer_sizes=(n_hidden[ind[0]], ),
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=random.randint(0, 1000))
reg.fit(x_train, y_train)
y_pred_test = reg.predict(x_test)
y_pred_train = reg.predict(x_train)
plot_learned_function(n_hidden[ind[0]], x_train, y_train, y_pred_train,
x_test, y_test, y_pred_test)
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
n_seeds = 10
n_neur = [1, 2, 3, 4, 6, 8, 12, 20, 40]
mse_train = np.zeros([np.size(n_neur), n_seeds])
mse_test = np.zeros([np.size(n_neur), n_seeds])
for h in range(np.size(n_neur)):
for s in range(n_seeds):
seed = np.random.randint(100)
reg = MLPRegressor(hidden_layer_sizes=(n_neur[h], ),
max_iter=5000,
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=seed)
reg.fit(x_train, y_train)
mse_train[h, s] = calculate_mse(reg, x_train, y_train)
mse_test[h, s] = calculate_mse(reg, x_test, y_test)
plot_mse_vs_neurons(mse_train, mse_test, n_neur)
sum_mse = mse_test.sum(axis=1)
ind_min = sum_mse.argmin()
reg = MLPRegressor(hidden_layer_sizes=(n_neur[ind_min], ),
max_iter=5000,
activation='logistic',
solver='lbfgs',
alpha=0,
random_state=np.random.randint(100))
reg.fit(x_train, y_train)
y_pred_test = reg.predict(x_test)
y_pred_train = reg.predict(x_train)
plot_learned_function(n_neur[ind_min], x_train, y_train, y_pred_train,
x_test, y_test, y_pred_test)
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model.
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
nh = [1, 2, 4, 6, 8, 12, 20, 40]
mse_all_train = np.zeros(shape=(8, 10))
mse_all_test = np.zeros(shape=(8, 10))
for i in range(0, 10):
for j in range(0, 8):
seed = np.random.randint(1, 100)
nn = MLPRegressor(activation='logistic',
solver='lbfgs',
max_iter=5000,
alpha=0,
hidden_layer_sizes=(nh[j], ),
random_state=seed)
nn.fit(x_train, y_train)
mse_train = calculate_mse(nn, x_train, y_train)
mse_test = calculate_mse(nn, x_test, y_test)
mse_all_train[j][i] = mse_train
mse_all_test[j][i] = mse_test
plot_mse_vs_neurons(mse_all_train, mse_all_test, nh)
nn = MLPRegressor(activation='logistic',
solver='lbfgs',
max_iter=5000,
alpha=0,
hidden_layer_sizes=(nh[2], ))
nn.fit(x_train, y_train)
y_pred_train = nn.predict(x_train)
y_pred_test = nn.predict(x_test)
plot_learned_function(nh[2], x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
pass
def ex_1_1_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 c)
Remember to set alpha to 0 when initializing the model
Use max_iter = 10000 and tol=1e-8
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
n_hidden_neurons_list = [1, 2, 3, 4, 6, 8, 12, 20, 40]
seeds = 10
mse = np.zeros((len(n_hidden_neurons_list), seeds, 2))
for i in range(len(n_hidden_neurons_list)):
for j in range(seeds):
regressor = MLPRegressor(
hidden_layer_sizes=(n_hidden_neurons_list[i], ),
activation='logistic',
solver='lbfgs',
alpha=0,
max_iter=10000,
random_state=j,
tol=1e-8)
regressor.fit(x_train, y_train)
# mse shape: [train_mses, test_mses]
mse[i][j] = calculate_mse(regressor, [x_train, x_test],
[y_train, y_test])
plot_mse_vs_neurons(mse[:, :, 0], mse[:, :, 1], n_hidden_neurons_list)
n_hidden = 40
regressor = MLPRegressor(hidden_layer_sizes=(n_hidden, ),
activation='logistic',
solver='lbfgs',
alpha=0,
max_iter=10000,
tol=1e-8)
regressor.fit(x_train, y_train)
y_pred_train = regressor.predict(x_train)
y_pred_test = regressor.predict(x_test)
plot_learned_function(n_hidden, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
## TODO
pass
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
## TODO
for i in [2, 8, 40]:
n_hidden_neurons = i
nn = MLPRegressor(activation='logistic', solver='lbfgs', max_iter=200, hidden_layer_sizes=(n_hidden_neurons,), alpha=0)
nn.fit(x_train, y_train)
predictions_test = nn.predict(x_test)
plot_learned_function(n_hidden_neurons, x_train, y_train, 0, x_test, y_test, predictions_test)
plt.show()
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
n_hidden = 40
trained_regressor = MLPRegressor(hidden_layer_sizes=(n_hidden, ),
activation='logistic',
solver='lbfgs',
alpha=0,
max_iter=200,
random_state=1000)
trained_regressor = trained_regressor.fit(x_train, y_train)
y_pred_train = trained_regressor.predict(x_train)
y_pred_test = trained_regressor.predict(x_test)
plot_learned_function(n_hidden, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
pass
def ex_1_1_a(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.1 a)
Remember to set alpha to 0 when initializing the model
:param x_train: The training dataset
:param x_test: The testing dataset
:param y_train: The training targets
:param y_test: The testing targets
:return:
"""
# declaring hidden layer neurons 2, 8, 40
hidden_layer_2 = 2
hidden_layer_8 = 8
hidden_layer_40 = 40
# declaring variables used in MLP-Regressor
activation_mode = 'logistic'
solver_mode = 'lbfgs'
alpha = 0
max_iter = 200
# declaring MLP-Regressor:
nn_2 = MLPRegressor(hidden_layer_sizes=(hidden_layer_2, ),
activation=activation_mode,
solver=solver_mode,
alpha=alpha,
max_iter=max_iter)
nn_8 = MLPRegressor(hidden_layer_sizes=(hidden_layer_8, ),
activation=activation_mode,
solver=solver_mode,
alpha=alpha,
max_iter=max_iter)
nn_40 = MLPRegressor(hidden_layer_sizes=(hidden_layer_40, ),
activation=activation_mode,
solver=solver_mode,
alpha=alpha,
max_iter=max_iter)
# train neural network using the regressor method fit
nn_2.fit(x_train, y_train)
nn_8.fit(x_train, y_train)
nn_40.fit(x_train, y_train)
# compute the output using the method predict
y_test_pred_2 = nn_2.predict(x_test)
y_train_pred_2 = nn_2.predict(x_train)
y_test_pred_8 = nn_8.predict(x_test)
y_train_pred_8 = nn_8.predict(x_train)
y_test_pred_40 = nn_40.predict(x_test)
y_train_pred_40 = nn_40.predict(x_train)
# plotting learned function
#def plot_learned_function(n_hidden, x_train, y_train, y_pred_train, x_test, y_test, y_pred_test):
plot_learned_function(hidden_layer_2, x_train, y_train, y_train_pred_2,
x_test, y_test, y_test_pred_2)
plot_learned_function(hidden_layer_8, x_train, y_train, y_train_pred_8,
x_test, y_test, y_test_pred_8)
plot_learned_function(hidden_layer_40, x_train, y_train, y_train_pred_40,
x_test, y_test, y_test_pred_40)
pass
def ex_1_2_c(x_train, x_test, y_train, y_test):
"""
Solution for exercise 1.2 c)
:param x_train:
:param x_test:
:param y_train:
:param y_test:
:return:
"""
## TODO
#pass
total_iter = 10000
epoch_iter = 20
epochs = total_iter // epoch_iter
n_neuro = 6
n_seeds = 10
"""
sequence = np.random.permutation(np.arange(0, np.size(y_train), 1))
x_train = x_train[sequence]
y_train = y_train[sequence]
SIZE = int(np.ceil(np.size(y_train) / 3))
x_val = x_train[:SIZE]
y_val = y_train[:SIZE]
x_train = x_train[SIZE:]
y_train = y_train[SIZE:]
"""
x_train, x_val, y_train, y_val = train_test_split(x_train,
y_train,
test_size=0.33)
mse_train = np.zeros([n_seeds, epochs])
mse_val = np.zeros([n_seeds, epochs])
mse_test = np.zeros([n_seeds, epochs])
#seeds = random.sample(range(1, 100), n_seeds)
seeds = np.zeros(n_seeds)
for s in range(n_seeds):
seed = s #np.random.randint(100)
seeds[s] = seed
reg = MLPRegressor(hidden_layer_sizes=(n_neuro, ),
activation='logistic',
solver='lbfgs',
alpha=1e-3,
random_state=seed,
warm_start=True,
max_iter=epoch_iter)
for ep in range(epochs):
reg.fit(x_train, y_train)
mse_train[s, ep] = calculate_mse(reg, x_train, y_train)
mse_val[s, ep] = calculate_mse(reg, x_val, y_val)
mse_test[s, ep] = calculate_mse(reg, x_test, y_test)
y_pred_test = reg.predict(x_test)
y_pred_train = reg.predict(x_train)
plot_learned_function(n_neuro, x_train, y_train, y_pred_train, x_test,
y_test, y_pred_test)
#min_val_index = np.argmin(mse_val, axis=1)
import pdb
pdb.set_trace()
min_val_index = np.unravel_index(mse_val.argmin(), mse_val.shape)
error_min_seed = seeds[min_val_index[0]]
error_min
#print(seeds)
#print(min_val_index)
print('Seed: ', error_min_seed)