def main():
print("-- XGBoost --")
from sklearn import preprocessing
from sklearn.preprocessing import LabelEncoder
with pyRAPL.Measurement('Read_data', output=csv_output):
df = pd.read_csv('/home/gabi/Teste/BaseSintetica/1k_5att.csv')
X = df.iloc[:, :-1].values
y = df.iloc[:, -1].values
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.4,
seed=2)
csv_output.save()
clf = XGBoost()
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)
def predict_value(self, x, tree=None):
with pyRAPL.Measurement('Predict_value', output=csv_output):
""" Do a recursive search down the tree and make a prediction of the data sample by the
value of the leaf that we end up at """
if tree is None:
tree = self.root
# If we have a value (i.e we're at a leaf) => return value as the prediction
if tree.value is not None:
return tree.value
# Choose the feature that we will test
feature_value = x[tree.feature_i]
# Determine if we will follow left or right branch
branch = tree.false_branch
if isinstance(feature_value, int) or isinstance(
feature_value, float):
if feature_value >= tree.threshold:
branch = tree.true_branch
elif feature_value == tree.threshold:
branch = tree.true_branch
# Test subtree
return self.predict_value(x, branch)
csv_output.save()
def fit(self, X, y, loss=None):
with pyRAPL.Measurement('Fit_1', output=csv_output):
""" Build decision tree """
self.one_dim = len(np.shape(y)) == 1
self.root = self._build_tree(X, y)
self.loss = None
csv_output.save()
def _split(self, y):
with pyRAPL.Measurement('Split', output=csv_output):
""" y contains y_true in left half of the middle column and
y_pred in the right half. Split and return the two matrices """
col = int(np.shape(y)[1] / 2)
y, y_pred = y[:, :col], y[:, col:]
return y, y_pred
csv_output.save()
def __init__(self, params):
self.params = params
try:
pyRAPL.setup()
self.rapl_enabled = True
self.meter_rapl = pyRAPL.Measurement("bar")
self.meter_rapl.begin()
except:
self.rapl_enabled = False
def fit(self, features, target):
with pyRAPL.Measurement('Fit', output=csv_output):
self.root = CART()
if (self.tree == 'cls'):
self.root._grow_tree(features, target, self.criterion)
else:
self.root._grow_tree(features, target, 'mse')
self.root._prune(self.prune, self.max_depth, self.min_criterion,
self.root.n_samples)
csv_output.save()
def _calculate_information_gain(self, y, y1, y2):
with pyRAPL.Measurement('Information_gain', output=csv_output):
# Calculate information gain
p = len(y1) / len(y)
entropy = calculate_entropy(y)
info_gain = entropy - p * \
calculate_entropy(y1) - (1 - p) * \
calculate_entropy(y2)
return info_gain
csv_output.save()
def _predict(self, d):
with pyRAPL.Measurement('Predict', output=csv_output):
if self.feature != None:
if d[self.feature] <= self.threshold:
return self.left._predict(d)
else:
return self.right._predict(d)
else:
return self.label
csv_output.save()
def _approximate_update(self, y):
with pyRAPL.Measurement('Approximate_update', output=csv_output):
# y split into y, y_pred
y, y_pred = self._split(y)
# Newton's Method
gradient = np.sum(y * self.loss.gradient(y, y_pred), axis=0)
hessian = np.sum(self.loss.hess(y, y_pred), axis=0)
update_approximation = gradient / hessian
return update_approximation
csv_output.save()
def _gain_by_taylor(self, y, y1, y2):
with pyRAPL.Measurement('Gain_by_taylor', output=csv_output):
# Split
y, y_pred = self._split(y)
y1, y1_pred = self._split(y1)
y2, y2_pred = self._split(y2)
true_gain = self._gain(y1, y1_pred)
false_gain = self._gain(y2, y2_pred)
gain = self._gain(y, y_pred)
return true_gain + false_gain - gain
csv_output.save()
def _majority_vote(self, y):
with pyRAPL.Measurement('Majority_vote', output=csv_output):
most_common = None
max_count = 0
for label in np.unique(y):
# Count number of occurences of samples with label
count = len(y[y == label])
if count > max_count:
most_common = label
max_count = count
return most_common
csv_output.save()
def _show_tree(self, depth, cond):
with pyRAPL.Measurement('Show_tree', output=csv_output):
base = ' ' * depth + cond
if self.feature != None:
print(base + 'if X[' + str(self.feature) + '] <= ' +
str(self.threshold))
self.left._show_tree(depth + 1, 'then ')
self.right._show_tree(depth + 1, 'else ')
else:
print(base + '{value: ' + str(self.label) + ', samples: ' +
str(self.n_samples) + '}')
csv_output.save()
def fit(self, X, y):
with pyRAPL.Measurement('Fit_3', output=csv_output):
y = to_categorical(y)
y_pred = np.zeros(np.shape(y))
for i in self.bar(range(self.n_estimators)):
tree = self.trees[i]
y_and_pred = np.concatenate((y, y_pred), axis=1)
tree.fit(X, y_and_pred)
update_pred = tree.predict(X)
y_pred -= np.multiply(self.learning_rate, update_pred)
csv_output.save()
def measure_power(self, label):
meter = pyRAPL.Measurement(label=label)
meter.begin()
time.sleep(self.MEASURE_TIME)
meter.end()
m_energy = meter._results.pkg[self.SOCKET] # micro-J
m_time = meter._results.duration # micro-s
power = m_energy / m_time # watts
return power
def _split_tree(self, features, target, criterion):
with pyRAPL.Measurement('Split_tree', output=csv_output):
features_l = features[features[:, self.feature] <= self.threshold]
target_l = target[features[:, self.feature] <= self.threshold]
self.left = CART()
self.left.depth = self.depth + 1
self.left._grow_tree(features_l, target_l, criterion)
features_r = features[features[:, self.feature] > self.threshold]
target_r = target[features[:, self.feature] > self.threshold]
self.right = CART()
self.right.depth = self.depth + 1
self.right._grow_tree(features_r, target_r, criterion)
csv_output.save()
def rapl_power(label, powertime, sockets):
meter = pyRAPL.Measurement(label=label)
while meter._results is None or meter._results.pkg is None:
meter.begin()
time.sleep(powertime)
meter.end()
results = {}
m_time = meter._results.duration # micro-s
for skt in sockets:
m_energy = meter._results.pkg[skt] # micro-J
results[skt] = m_energy / m_time # watts
return results
def before(self):
try:
pyRAPL.setup()
self.meter = pyRAPL.Measurement('bar')
self.meter.begin()
self.successful = True
except FileNotFoundError:
logging.warning(
"RAPL file not found. Perhaps you are using a platform that does not support RAPL (for example Windows)"
)
self.successful = False
except PermissionError:
logging.warning(
"PermissionError occured while reading RAPL file. Fix with \"sudo chmod -R a+r /sys/class/powercap/intel-rapl\""
)
self.successful = False
def _calc_impurity(self, criterion, target):
with pyRAPL.Measurement('Calc_impurity', output=csv_output):
if criterion == 'gini':
return 1.0 - sum([(float(len(target[target == c])) /
float(target.shape[0]))**2.0
for c in np.unique(target)])
elif criterion == 'mse':
return np.mean((target - np.mean(target))**2.0)
else:
entropy = 0.0
for c in np.unique(target):
p = float(len(target[target == c])) / target.shape[0]
if p > 0.0:
entropy -= p * np.log2(p)
return entropy
csv_output.save()
def _grow_tree(self, features, target, criterion='gini'):
with pyRAPL.Measurement('Grow_tree', output=csv_output):
self.n_samples = features.shape[0]
if len(np.unique(target)) == 1:
self.label = target[0]
return
best_gain = 0.0
best_feature = None
best_threshold = None
if criterion in {'gini', 'entropy'}:
self.label = max([(c, len(target[target == c]))
for c in np.unique(target)],
key=lambda x: x[1])[0]
else:
self.label = np.mean(target)
impurity_node = self._calc_impurity(criterion, target)
for col in range(features.shape[1]):
feature_level = np.unique(features[:, col])
thresholds = (feature_level[:-1] + feature_level[1:]) / 2.0
for threshold in thresholds:
target_l = target[features[:, col] <= threshold]
impurity_l = self._calc_impurity(criterion, target_l)
n_l = float(target_l.shape[0]) / self.n_samples
target_r = target[features[:, col] > threshold]
impurity_r = self._calc_impurity(criterion, target_r)
n_r = float(target_r.shape[0]) / self.n_samples
impurity_gain = impurity_node - (n_l * impurity_l +
n_r * impurity_r)
if impurity_gain > best_gain:
best_gain = impurity_gain
best_feature = col
best_threshold = threshold
self.feature = best_feature
self.gain = best_gain
self.threshold = best_threshold
self._split_tree(features, target, criterion)
csv_output.save()
def predict(self, X):
with pyRAPL.Measurement('Predict_2', output=csv_output):
y_pred = None
# Make predictions
for tree in self.trees:
# Estimate gradient and update prediction
update_pred = tree.predict(X)
if y_pred is None:
y_pred = np.zeros_like(update_pred)
y_pred -= np.multiply(self.learning_rate, update_pred)
# Turn into probability distribution (Softmax)
y_pred = np.exp(y_pred) / np.sum(
np.exp(y_pred), axis=1, keepdims=True)
# Set label to the value that maximizes probability
y_pred = np.argmax(y_pred, axis=1)
return y_pred
csv_output.save()
def predict(self, X):
with pyRAPL.Measurement('Predict_2', output=csv_output):
y_preds = np.empty((X.shape[0], len(self.trees)))
# Let each tree make a prediction on the data
for i, tree in enumerate(self.trees):
# Indices of the features that the tree has trained on
idx = tree.feature_indices
# Make a prediction based on those features
prediction = tree.predict(X[:, idx])
y_preds[:, i] = prediction
y_pred = []
# For each sample
for sample_predictions in y_preds:
# Select the most common class prediction
y_pred.append(
np.bincount(sample_predictions.astype('int')).argmax())
return y_pred
csv_output.save()
def classification_example():
with pyRAPL.Measurement('Read_data', output=csv_output):
df = pd.read_csv('/home/gabrieli/Documentos/BaseSintetica/1k_5att.csv')
#df = df.apply(LabelEncoder().fit_transform)
X = df.iloc[:, :-1].values
y = df.iloc[:, -1].values
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.4)
csv_output.save()
cls = CART(tree='cls', criterion='gini', prune='depth')
cls.fit(X_train, y_train)
cls.print_tree()
pred = cls.predict(X_test)
print("Accuracy:", accuracy_score(y_test, pred))
def print_tree(self, tree=None, indent=" "):
with pyRAPL.Measurement('Print_tree', output=csv_output):
""" Recursively print the decision tree """
if not tree:
tree = self.root
# If we're at leaf => print the label
if tree.value is not None:
print(tree.value)
# Go deeper down the tree
else:
# Print test
print("%s:%s? " % (tree.feature_i, tree.threshold))
# Print the true scenario
print("%sT->" % (indent), end="")
self.print_tree(tree.true_branch, indent + indent)
# Print the false scenario
print("%sF->" % (indent), end="")
self.print_tree(tree.false_branch, indent + indent)
csv_output.save()
def main():
print("-- Classification Tree --")
with pyRAPL.Measurement('Read_data', output=csv_output):
dataset = pd.read_csv('/home/gabi/Teste/BaseSintetica/1k_5att.csv')
X = dataset.iloc[:, 0:5].values
y = dataset.iloc[:, 5].values
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.4)
csv_output.save()
clf = ClassificationTree()
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)
def test_context_measure(fs_one_socket):
"""
Test to measure the energy consumption of a function using the Measurement class
- launch the measure
- write a new value to the RAPL power measurement api file
- launch a function
- end the measure
Test if:
- the energy consumption measured is the delta between the first and the last value in the RAPL power measurement
file
"""
pyRAPL.setup()
out = dummyOutput()
with pyRAPL.Measurement('toto', output=out):
measurable_function(1)
assert out.data.pkg == [(POWER_CONSUMPTION_PKG - PKG_0_VALUE)]
assert out.data.dram == [(POWER_CONSUMPTION_DRAM - DRAM_0_VALUE)]
def test_nomal_measure_bench(fs_one_socket):
"""
Test to measure the energy consumption of a function using the Measurement class
- launch the measure
- write a new value to the RAPL power measurement api file
- launch a function
- end the measure
Test if:
- the energy consumption measured is the delta between the first and the last value in the RAPL power measurement
file
"""
pyRAPL.setup()
measure = pyRAPL.Measurement('toto')
measure.begin()
measurable_function(1)
measure.end()
assert measure.result.pkg == [(POWER_CONSUMPTION_PKG - PKG_0_VALUE)]
assert measure.result.dram == [(POWER_CONSUMPTION_DRAM - DRAM_0_VALUE)]
def regression_example():
with pyRAPL.Measurement('Read_data', output=csv_output):
df = pd.read_csv('/home/gabrieli/Documentos/BaseSintetica/1k_5att.csv')
#df = df.apply(LabelEncoder().fit_transform)
X = df.iloc[:, :-1].values
y = df.iloc[:, -1].values
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.4)
csv_output.save()
# Fit regression model
reg = CART(tree='reg', criterion='mse', prune='depth')
reg.fit(X_train, y_train)
reg.print_tree()
pred = reg.predict(X_test)
mse = mean_squared_error(y_test, pred)
print("MSE:", mse)
def _prune(self, method, max_depth, min_criterion, n_samples):
with pyRAPL.Measurement('Prune', output=csv_output):
if self.feature is None:
return
self.left._prune(method, max_depth, min_criterion, n_samples)
self.right._prune(method, max_depth, min_criterion, n_samples)
pruning = False
if method == 'impurity' and self.left.feature is None and self.right.feature is None:
if (self.gain * float(self.n_samples) /
n_samples) < min_criterion:
pruning = True
elif method == 'depth' and self.depth >= max_depth:
pruning = True
if pruning is True:
self.left = None
self.right = None
self.feature = None
csv_output.save()
def fit(self, X, y):
with pyRAPL.Measurement('Fit_3', output=csv_output):
n_features = np.shape(X)[1]
# If max_features have not been defined => select it as
# sqrt(n_features)
if not self.max_features:
self.max_features = int(math.sqrt(n_features))
# Choose one random subset of the data for each tree
subsets = get_random_subsets(X, y, self.n_estimators)
for i in self.progressbar(range(self.n_estimators)):
X_subset, y_subset = subsets[i]
# Feature bagging (select random subsets of the features)
idx = np.random.choice(range(n_features),
size=self.max_features,
replace=True)
# Save the indices of the features for prediction
self.trees[i].feature_indices = idx
# Choose the features corresponding to the indices
X_subset = X_subset[:, idx]
# Fit the tree to the data
self.trees[i].fit(X_subset, y_subset)
csv_output.save()
def _build_tree(self, X, y, current_depth=0):
with pyRAPL.Measurement('Build_tree', output=csv_output):
""" Recursive method which builds out the decision tree and splits X and respective y on the feature of X which (based on impurity) best separates the data"""
largest_impurity = 0
best_criteria = None # Feature index and threshold
best_sets = None # Subsets of the data
# Check if expansion of y is needed
if len(np.shape(y)) == 1:
y = np.expand_dims(y, axis=1)
# Add y as last column of X
Xy = np.concatenate((X, y), axis=1)
n_samples, n_features = np.shape(X)
if n_samples >= self.min_samples_split and current_depth <= self.max_depth:
# Calculate the impurity for each feature
for feature_i in range(n_features):
# All values of feature_i
feature_values = np.expand_dims(X[:, feature_i], axis=1)
unique_values = np.unique(feature_values)
# Iterate through all unique values of feature column i and
# calculate the impurity
for threshold in unique_values:
# Divide X and y depending on if the feature value of X at index feature_i
# meets the threshold
Xy1, Xy2 = divide_on_feature(Xy, feature_i, threshold)
if len(Xy1) > 0 and len(Xy2) > 0:
# Select the y-values of the two sets
y1 = Xy1[:, n_features:]
y2 = Xy2[:, n_features:]
# Calculate impurity
impurity = self._impurity_calculation(y, y1, y2)
# If this threshold resulted in a higher information gain than previously
# recorded save the threshold value and the feature
# index
if impurity > largest_impurity:
largest_impurity = impurity
best_criteria = {
"feature_i": feature_i,
"threshold": threshold
}
best_sets = {
"leftX":
Xy1[:, :n_features], # X of left subtree
"lefty":
Xy1[:, n_features:], # y of left subtree
"rightX":
Xy2[:, :n_features], # X of right subtree
"righty":
Xy2[:, n_features:] # y of right subtree
}
if largest_impurity > self.min_impurity:
# Build subtrees for the right and left branches
true_branch = self._build_tree(best_sets["leftX"],
best_sets["lefty"],
current_depth + 1)
false_branch = self._build_tree(best_sets["rightX"],
best_sets["righty"],
current_depth + 1)
return DecisionNode(feature_i=best_criteria["feature_i"],
threshold=best_criteria["threshold"],
true_branch=true_branch,
false_branch=false_branch)
# We're at leaf => determine value
leaf_value = self._leaf_value_calculation(y)
return DecisionNode(value=leaf_value)
csv_output.save()