def main():
print('--- Adaboost ---')
data = datasets.load_digits()
X, y = data.data, data.target
digit1 = 1
digit2 = 8
idx = np.append(np.where(y == digit1)[0], np.where(y == digit2)[0])
y = data.target[idx]
y[y == digit1] = 1
y[y == digit2] = -1
X = data.data[idx]
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)
clf = Adaboost(n_estimators=5)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
acc = accuracy_score(y_pred, y_test)
clf_tree = ClassificationTree()
clf_tree.fit(X_train, y_train)
y_pred_tree = clf_tree.predict(X_test)
acc_tree = accuracy_score(y_pred, y_test)
print("Adaboost_Accuracy:", acc)
print("Tree_Accuracy:", acc_tree)
def fit(self, X, y):
n_samples, n_features = X.shape[0], X.shape[1]
# 初始化各样本的权重
w = np.full(n_samples, (1 / n_samples))
# 储存每个分类器
self.clfs = []
for i in tqdm(range(self.n_clfs)):
# 实例化一个分类器
clf = ClassificationTree()
# 训练
clf.fit(X, y)
# 得到训练结果
y_pred = clf.predict(X)
# 计算训练误差
print(accuracy_score(y, y_pred))
error = sum(w[y != y_pred])
# 对于错误率大于0.5的分类器，因为是二分类问题(Adaboost只能解决二分类问题)，我们可以翻转
# 分类器的预测结果来使得其错误率为1-error>0.5
print(error)
if error > 0.5:
self.polarity[i] = -1
y_pred *= -1
error = 1 - error
self.alphas[i] = 0.5 * np.log((1.0 - error) / (error + 1e-10))
predictions = np.array(self.polarity[i] * y_pred)
w *= np.exp(-self.alphas[i] * y * predictions)
w /= sum(w)
self.clfs.append(clf)
def main():
data = datasets.load_digits()
X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4, seed=2)
clf = RandomForest(n_estimators=10)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print ("Accuracy:", accuracy)
def main():
print ("-- XGBoost --")
data = datasets.load_iris()
X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1, seed=3)
clf = XGBoost(n_estimators=20)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print ("Accuracy:", accuracy)
def main():
print("-- Classification Tree --")
data = datasets.load_iris()
X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4)
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 main():
data = datasets.load_digits()
X = data.data
y = data.target
y = to_categorical(y.astype("int"))
n_hidden = 512
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4)
clf = MultilayerPerceptron(n_hidden)
clf.fit(X_train, y_train)
y_pred = np.argmax(clf.predict(X_test), axis=1)
y_test = np.argmax(y_test, axis=1)
accuracy = accuracy_score(y_test, y_pred)
print(f"Accuracy: {accuracy}")
Plot().plot_in_2d(X_test, y_pred, title="perceptron", accuracy=accuracy, legend_labels=np.unique(y))
def main_classifier():
print("-- Gradient Boosting Classification --")
data = datasets.load_iris()
X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4)
print(y_train.shape)
clf = GradientBoostingClassifier(n_estimators=10)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)
def main():
data = datasets.load_digits()
X = normalize(data.data)
y = data.target
# data preprocess: One-hot encoding of nominal y-values
y = to_categorical(y)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33)
clf = Perceptron(n_iterations=5000, learning_rate=0.001, loss=CrossEntropy, activation_function=Sigmoid)
clf.fit(X_train, y_train)
y_pred = np.argmax(clf.predict(X_test), axis=1)
y_test = np.argmax(y_test, axis=1)
accuracy = accuracy_score(y_test, y_pred)
print(f"Accuracy: {accuracy}")
Plot().plot_in_2d(X_test, y_pred, title="perceptron", accuracy=accuracy, legend_labels=np.unique(y))
def evaluate(self, sess, data_manager, id_to_tag): # 评价slot和intent的分数
"""
:param sess: session to run the model
:param data: list of data
:param id_to_tag: index to tag name
:param id_to_intent: index to intent name
:return: evaluate result
"""
slot_results = []
itent_results = []
trans = self.trans.eval()
for batch in data_manager.iter_batch():
strings = batch[0]
tags = batch[-2] # 真实的slot标签
intents = np.asarray(batch[-1])[:, 1] # 真实的intents标签
lengths, scores_slot, intent_idx, intent_rank = self.run_step(
sess, False, batch)
batch_paths = self.decode(scores_slot, lengths,
trans) # viterbi算法求出最佳路径
for i in range(len(strings)):
result = []
string = strings[i][:lengths[i]]
gold = iobes_iob(
[id_to_tag[int(x)] for x in tags[i][:lengths[i]]])
pred = iobes_iob(
[id_to_tag[int(x)] for x in batch_paths[i][:lengths[i]]])
for char, gold, pred in zip(string, gold, pred):
result.append(" ".join([char, gold, pred]))
slot_results.append(result)
intent_acc = accuracy_score(intents, intent_idx)
itent_results.append(intent_acc)
return slot_results, itent_results
def main():
data = datasets.load_digits()
X = data.data
y = data.target
digit1 = 1
digit2 = 8
idx = np.append(np.where(y == digit1)[0], np.where(y == digit2)[0])
y = data.target[idx]
# Change labels to {-1, 1}
y[y == digit1] = -1
y[y == digit2] = 1
X = data.data[idx]
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)
# Adaboost classification with 5 weak classifiers
clf = Adaboost_1(n_clfs=5)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)
import numpy as np
from sklearn import datasets
import matplotlib.pyplot as plt
from deep.activations_functions import Sigmoid
from deep.loss_functions import CrossEntropy
from deep.perceptron import Perceptron
from utils.utils import accuracy_score
np.random.seed(0)
X, y = datasets.make_moons(200, noise=0.20)
clf = Perceptron(n_iterations=5000, learning_rate=0.001, loss=CrossEntropy, activation_function=Sigmoid)
clf.fit(X, y)
x_min, x_max = X[:, 0].min() - .5, X[:, 0].max() + .5
y_min, y_max = X[:, 1].min() - .5, X[:, 1].max() + .5
h = 0.01
xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
tmp = np.c_[xx.ravel(), yy.ravel()]
Z = clf.predict(tmp)
y_pred = np.argmax(Z, axis=1)
accuracy = accuracy_score(y, y_pred)
print(f"Accuracy: {accuracy}")
plt.contourf(xx, yy, y_pred.reshape(xx.shape), cmap=plt.cm.get_cmap("Spectral"))
plt.scatter(X[:, 0], X[:, 1], c=y, cmap=plt.cm.get_cmap("Spectral"))
plt.show()
def accuracy(self, y, p):
return accuracy_score(np.argmax(y, axis=1), np.argmax(p, axis=1))
self.parameters[c][feature_i]['var'], feature_value)
posterior *= likelihood
else:
posterior *= self.parameters[c][feature_i]
# 储存x为类别c的概率
posteriors.append(posterior)
# 返回概率最大的类别
return self.classes[np.argmax(posteriors)]
def predict(self, X_test):
y_pred = [self.get_label(x) for x in X_test]
return y_pred
if __name__ == '__main__':
print("-- Navie-Bayes --")
data = datasets.load_iris()
X = data.data
y = data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5)
clf = Navie_Bayes()
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy:", accuracy)
def main(args):
torch.manual_seed(17) # Randomly seed PyTorch
# Load train, dev, test iterators with auto-batching and pretrained vectors
(train, dev, test, code_vecs,
comm_vecs) = utils.load_all_data(args.batch_size, args.corpus)
# Create model
model = cl.CodeCommClassifier(code_vecs,
comm_vecs,
gpu=args.use_gpu,
model_type=args.model)
# Load a saved model
if args.load != "":
model.load_state_dict(torch.load(args.load))
# Discover available CUDA devices and use DataParallelism if possible
devices = list(range(torch.cuda.device_count()))
if len(devices) > 1:
print("Using {} GPUs!!".format(len(devices)))
model = nn.DataParallel(model)
# Send model to GPU for processing
if args.use_gpu:
model.cuda()
# Do training
if args.epochs > 0:
print("Training {} model w/ batch_sz={} on {} dataset".format(
args.model, args.batch_size, args.corpus))
# Adam optimizer works, SGD fails to train the network for any batch size
optimizer = optim.Adam(model.parameters(), args.learning_rate)
class_weights = torch.tensor([10.])
if args.use_gpu:
class_weights = class_weights.cuda()
for epoch in range(args.epochs):
model.train() # Set the model to training mode
with tqdm(total=len(train),
desc="Epoch {}/{}".format(epoch + 1,
args.epochs)) as pbar:
total_loss = 0
for b_idx, batch in enumerate(train):
# model.zero_grad()
optimizer.zero_grad() # Clear current gradient
# Get the label into a tensor for loss prop
truth = torch.autograd.Variable(batch.label).float()
if args.use_gpu:
truth = truth.cuda()
# Run the model using the batch
outputs = model((batch.code, batch.comm))
# Get loss from log(softmax())
loss = F.binary_cross_entropy_with_logits(
outputs.view(-1),
truth.view(-1),
pos_weight=class_weights)
loss.backward() # Propagate loss
optimizer.step() # Update the optimizer
new_loss = loss.item()
total_loss += new_loss
curr_loss = total_loss / (b_idx + 1)
pbar.set_postfix(batch_loss=curr_loss)
pbar.update()
# Check DEV accuracy after every epoch
scores = score_dataset(model, dev)
acc = utils.accuracy_score(scores)
sys.stdout.write("Epoch {} -- dev acc: {}%\n".format(
epoch + 1, acc))
# Save the model weights
if args.save != "":
torch.save(model.state_dict(), args.save)
# Save the DEV set evaluations
if args.eval_dev:
print(
"Evaluating Dev for {} model w/ batch_sz={} on {} dataset".format(
args.model, args.batch_size, args.corpus))
scores = score_dataset(model, dev)
dev_score_path = utils.CODE_CORPUS / "results" / "{}_{}_{}_{}_gpus_scores_dev.pkl".format(
args.model, args.batch_size, args.corpus, len(devices))
utils.save_scores(scores, dev_score_path)
# Save the TEST set evaluations
if args.eval_test:
print(
"Evaluating Dev for {} model w/ batch_sz={} on {} dataset".format(
args.model, args.batch_size, args.corpus))
scores = score_dataset(model, test)
test_score_path = utils.CODE_CORPUS / "results" / "{}_{}_{}_{}_gpus_scores_test.pkl".format(
args.model, args.batch_size, args.corpus, len(devices))
utils.save_scores(scores, test_score_path)
